# Beyond APIs: How AI Web Agents Are Automating the "Long Tail" of Knowledge Work

Arjun Chintapalli · Bhavani Kalisetty

AI Engineer Summit 2025

[Talk page](https://ai.engineer/talks/kjSGc7uwDo8-beyond-apis-ai-web-agents-are-automating) · [Video](https://www.youtube.com/watch?v=kjSGc7uwDo8)

## Beyond APIs: How AI Web Agents Are Automating the "Long Tail" of Knowledge Work

A browser agent can combine page interaction, structured extraction and API calls, turning scattered tabs into spreadsheet data, research inputs and actions.

### When the browser becomes the bottleneck

[0:01](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1s)

How much work remains after you find the information you need on a website? Someone still has to copy it into Google Sheets or a CRM, reconcile it with data from an API, and repeat the process on the next page. [rtrvr.ai](https://rtrvr.ai/), pronounced Retriever, brings a natural-language agent into a Chrome side panel to combine browser actions, structured extraction and API calls. Arjun Chintapalli and Bhavani Kalisetty demonstrate how those capabilities fit together.

The bottleneck is more than manual entry. Outsourced scraping introduces cost and reliability concerns; traditional RPA bots can break when a website changes. Meanwhile, information available only through a website remains separate from information available through an API. An agent operating alongside normal browsing offers a way to connect those sources and move results into a working spreadsheet.

![Orange slide titled “Your Browser is the Bottleneck.” Four text blocks describe workflow problems, with the presenter visible at right.](https://ai.engineer/api/talk-essay-media/165288e919d68168478e7cf7767e7cc9976b672e1aa1a2c618dbc2903007fdc3)

Four browser workflow bottlenecks: manual data entry, offshore scraping, brittle RPA bots, and data silos.

### Check page state, then act

[1:38](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=98s)

The first task is small but conditional: find the Latent Space podcast page on LinkedIn and follow it, unless the account already follows it. After the instruction is submitted, the agent fills the search field and inspects the page. It recognizes that the account is already following the page, so the requested outcome requires no new follow action. **The instruction describes a desired state, including when to leave the page unchanged.**

The next example moves from interaction to structured output. A request specifies data to extract for every article on a page, and the export operation writes the results to Google Sheets. These capabilities can be composed: first perform a task on the page, then extract the resulting data. Chintapalli estimates that this particular page extraction probably cost less than one cent; the demonstration does not establish a measured cost or a general per-page price.

### Spread the work across tabs

[3:09](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=189s)

A spreadsheet can also supply the input. Given a Google Sheet column of URLs, the agent can open those pages, interact with them and extract their contents. In the arXiv example, the request asks for specified fields from the first five PDFs in a search. The agent splits the work into subtasks, opens new tabs and processes them independently and simultaneously before writing results to Sheets. Applying to LinkedIn jobs is offered as another possible use of the same tab-based action pattern.

For product comparison, the user instead selects three Amazon product tabs. Leaving the prompt empty lets the agent infer likely comparison fields and collect them into a spreadsheet. Explicit instructions are still useful when the desired output goes beyond those inferred fields: the demonstration also extracts source image URLs so the products' images can be compared alongside their other data.

### Change the review order before extracting

[4:40](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=280s)

The Amazon review example makes the dependency between action and extraction concrete. The pages initially show top reviews, but the request asks for the most recent review. Extracting the current contents immediately would answer the wrong question. The workflow is:

1. Select the product tabs to process.
2. In each tab, change the review dropdown from top reviews to most recent.
3. Extract the details of the most recent review from the updated page.

The agent performs the actions across the selected tabs in the background. Chintapalli points out the dropdown changing on the pages, then reports that the recent reviews have been extracted. **The page state is part of the extraction task**, not merely a source to read once.

### Use selected documents as research context

[6:10](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=370s)

The next workflow starts with design documents already open in the browser. To prepare a quick summary for a colleague, the user selects the documents and requests their key points and summaries. Selection defines which sources the agent should read; the input can mix Google Docs, PDFs and Google Sheets. The resulting response distinguishes the individual documents, including one concerning LLMs with memory. This is a synthesis task over a chosen collection, rather than an instruction to search the web indiscriminately.

![Google Docs displays the end of an AI web agent design document, while the sidebar shows an extraction request and a long response headed “Task Response.”](https://ai.engineer/api/talk-essay-media/3c4c612a1ee1017b2d4f771c85a51d2f5935a3e3e6c2f5b06f2be9e2a0ff8e59)

A document summary appears in the rtrvr.ai sidebar beside a Google Doc.

### Let missing fields guide navigation

[7:08](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=428s)

Company research extends the task beyond the pages initially selected. Kalisetty asks for strategy, features and pricing across Retriever, [Bardeen](https://www.bardeen.ai/) and [Browse AI](https://www.browse.ai/). The agent first generates an extraction schema from the request. Starting from the companies' homepages, it then navigates to pricing pages to locate information missing from the initial view. The requested fields therefore guide both the output structure and the browsing work needed to populate it.

Kalisetty positions this as an early deep-research capability relative to [OpenAI's Operator](https://openai.com/index/introducing-operator/). The practical feature in the demonstration is control over the research inputs: users can select documents they have access to and specify the URLs or websites the agent may visit. The collected company information is written to Google Sheets.

### Recover from an unsuitable starting URL

[8:27](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=507s)

The stock-research example begins with rows in a Google Sheet. The request asks for a P/E ratio from Yahoo Finance, revenue for the last two years and a new field computing revenue growth from those values. This combines three operations: read the existing stock records, retrieve additional financial data and derive a value from the retrieved numbers.

One supplied URL points to a company's website rather than Yahoo Finance. The agent recognizes that the starting page does not match the requested source and navigates to Yahoo Finance to continue the extraction into Sheets. The recovery illustrates how the task description can override an unsuitable starting URL. It does not establish the accuracy of the financial values or demonstrate an investment decision.

### Turn spreadsheet rows into WhatsApp messages

[9:42](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=582s)

Browser interaction can feed an API operation rather than end with an export. Retriever's **dynamic function calling** lets users provide information about a tool instead of depending entirely on a fixed catalog of vendor-built connectors. The broad promise is integration with arbitrary third-party APIs; the concrete example here uses a configured WhatsApp messaging tool.

The selected spreadsheet contains two customers' phone numbers. Kalisetty asks the agent to send messages to those numbers. It identifies each number and invokes the WhatsApp tool, then reports that all messages were sent. Kalisetty opens WhatsApp to check the result and points to a received test message timestamped 2:32, matching the demonstration's local time. That visible receipt supports delivery to the displayed test number, separately from the agent's broader success report.

![WhatsApp Web shows the Test Number conversation with repeated “Hello World” messages; the rtrvr.ai sidebar displays “All messages sent.”](https://ai.engineer/api/talk-essay-media/678a9ffba7ef377a66602bb161869beca53574db7cd4568240c8659ae48ccdef)

WhatsApp displays test messages alongside the agent’s message-sending report.

### Generate a representation of the data

[11:04](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=664s)

The messaging example suggests broader social-communication workflows across Instagram, Facebook and WhatsApp, though those additional deliveries are not demonstrated. The next demonstrated tool moves in a different direction: a graph bot, described as a mini agent inside Retriever Agent Studio, takes existing data and generates an analysis graph on demand. The same workflow that collects structured information can therefore also produce a representation intended for a person to inspect.

### Read webpage text inside the user's session

[12:05](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=725s)

The architecture discussion separates how an agent observes a page from where its browser runs. Kalisetty contrasts Retriever's text-based approach with the screenshot-based interaction used by the launch versions of Operator and [**Claude** computer use](https://www.anthropic.com/news/developing-computer-use), and describes Google Mariner as hybrid. Screenshot-based observation captures the page visually as the agent acts or scrolls; Retriever instead supplies webpage text to the model. The speakers attribute lower hallucination and cost to that choice, but provide no controlled comparison establishing those advantages.

The second distinction is the browser session itself:

| Deployment | Relevant context | Infrastructure concern raised in the talk |
|---|---|---|
| Cloud browser | A separate session may show different content | Proxy routing and IP infrastructure |
| Local extension | The user's existing browser session | Uses the browser already on the user's machine |

Kalisetty names Browserbase while discussing cloud-hosted browsers and argues that maintaining the necessary network routing adds expense. The useful distinction is session context and deployment overhead; the talk does not measure their relative costs.

The extension can process active and background tabs while using content the user can already access. The speakers state that Retriever does not require users to share passwords and does not store them. Login-walled and subscribed content is available through the existing authenticated session; this is not a demonstrated universal bypass of paywalls or Cloudflare protections. Nor must every cloud agent expose passwords to its model: Operator documented manual takeover for login without collecting or screenshotting information entered during that takeover.

### Use text to work across background tabs

[14:26](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=866s)

Chintapalli connects text-based observation to the earlier multi-tab demonstrations. With the page text directly in context, the model can use that content without requiring a screenshot of each background tab. In the described implementation, those tabs need not be visually rendered for the agent to act on them. This supports both independent work in parallel and actions informed by context from multiple tabs.

That is a specific advantage for background-tab interaction inside one local browser, not a universal prohibition on parallel vision-based agents. Operator's launch documentation also described simultaneous tasks in separate conversations. The architectural question is how the system obtains observations and isolates concurrent work, rather than whether vision alone makes concurrency impossible.

### Decompose repetitive work into smaller tasks

[15:12](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=912s)

The product focus is repetitive knowledge work rather than consumer tasks such as booking flights or restaurants. The client-side Chrome extension is presented as a way to reduce hosting overhead and reach local or login-walled sites through the user's browser. This is the architecture shown in the recording; today's Retriever offering also includes cloud execution and other access methods.

Task decomposition is the corresponding reliability strategy. Instead of keeping a long sequence of actions on one tab, Retriever distributes subtasks to new tabs. Chintapalli argues that this reduces failures and avoids compounding errors, although no benchmark establishes a lower failure rate. For integrations, the parallel scaling strategy is to let users define, configure and share function calls, expanding the available tools without requiring the vendor to build every connector itself.

### Build datasets across volunteers' browsers

[16:31](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=991s)

The closing proposal extends extraction beyond one person's workflow: an AI-powered exchange in which people collaborate through their local laptops to construct datasets. Consider a collection of local government events across the San Francisco Bay Area. Gathering it could require visiting hundreds or thousands of websites, making centralized collection costly. Chintapalli proposes that volunteers contribute extraction work through their extensions, sharing the effort of building otherwise difficult-to-assemble datasets. This exchange is a long-term goal, not a deployed capability demonstrated in the talk.

The recording closes with an invitation to download the extension from rtrvr.ai or use the on-screen QR code. The immediate entry point is an agent working beside the user in the browser; the proposed destination is collaborative collection across many such browsers.

![Orange slide reading “rtrvr.ai: Unlock the Full Potential of Your Browser!” above a large QR code, with contact and support addresses at lower right.](https://ai.engineer/api/talk-essay-media/2980d0e7ef11612dcd234e2262f5f5866c75a1a99c3b19fe3635995b6966710d)

The closing rtrvr.ai slide presents a QR code and contact information.

## Resources

- [Retriever AI](https://rtrvr.ai/) — Product homepage with access to the Chrome extension and current automation offerings.
- [Retriever AI tool-calling guide](https://rtrvr.ai/docs/tool-calling) — Current guide to custom tools, authenticated webpage execution and spreadsheet-driven tool calls.
- [Introducing Operator](https://openai.com/index/introducing-operator/) — Operator's launch explanation of screenshot-based browser interaction, simultaneous tasks and user takeover.
- [Developing a computer use model](https://www.anthropic.com/news/developing-computer-use) — Anthropic's explanation of screenshot interpretation, cursor control and early computer-use limitations.
- [Bardeen](https://www.bardeen.ai/) — Web scraping and workflow automation for lead research, enrichment and data export.
- [Browse AI](https://www.browse.ai/) — No-code tools for extracting and monitoring website data.

## Slides

### Extracting arXiv paper details

The agent opens the first five arXiv PDFs in separate tabs and writes extracted fields to a sheet. Four rows show titles and authors; one shows URLs but no visible title or author.

[3:28](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=208s) [An arXiv artificial intelligence paper list beside a request to extract details from the first five PDFs.](https://assets.ai.engineer/sites/aiecode2025/sha256/12ba468e0b32d14e0c1f3b10b079421f8fc82ec8158ed3603d5a04b227a7cfaa.webp)

Artificial Intelligence
Authors and titles for recent submissions

Fri, 14 Feb 2025
Thu, 13 Feb 2025
Wed, 12 Feb 2025
Tue, 11 Feb 2025
Mon, 10 Feb 2025

See today's new changes
Total of 901 entries: 1–50 51–100 101–150 151–200 ... 901–901
Showing up to 50 entries per page: fewer | more | all

Fri, 14 Feb 2025 (showing first 50 of 159 entries)

[1] arXiv:2502.09601 [pdf, html, other]
CoT-Valve: Length-Compressible Chain-of-Thought Tuning
Xinyin Ma, Guangnian Wan, Runpeng Yu, Gongfan Fang, Xinchao Wang
Comments: Work in progress. Code will be released at this https URL
Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)

[2] arXiv:2502.09596 [pdf, html, other]
KIMAs: A Configurable Knowledge Integrated Multi-Agent System
Zitao Li, Fei Wei, Yuexiang Xie, Dawei Gao, Weirui Kuang, Zhijian Ma, Bingchen Qian, Yaliang Li, Bolin Ding
Subjects: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)

[3] arXiv:2502.09565 [pdf, html, other]
MDCrow: Automating Molecular Dynamics Workflows with Large Language Models
Quintina Campbell, Sam Cox, Jorge Medina, Brittany Watterson, Andrew D. White
Subjects: Artificial Intelligence (cs.AI); Chemical Physics (physics.chem-ph)

[4] arXiv:2502.09560 [pdf, html, other]
EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents
Rui Yang, Hanyang Chen, Junyu Zhang, Mark Zhao, Cheng Qian, Kangrui Wang, Qineng Wang, Teja Venkat Koripella, Marziyeh Movahedi, Manling Li, Heng Ji, Huan Zhang, Tong Zhang
Comments: 51 pages
Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)

[5] arXiv:2502.09432 [pdf, html, other]
Dual Formulation for Non-Rectangular Lp Robust Markov Decision Processes
Navdeep Kumar, Adarsh Gupta, Maxence Mohamed Elfatihi, Giorgia Ramponi, Kfir Yehuda Levy, Shie Mannor
Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)

[6] arXiv:2502.09378 [pdf, html, other]
A Deep Inverse-Mapping Model for a Flapping Robotic Wing
Hadar Sharvit, Raz Karl, Tsevi Beatus
Comments: Accepted to ICLR 2025. 10 Pages 5 figures + 2 figures in appendix
Subjects: Artificial Intelligence (cs.AI); Robotics (cs.RO)

[7] arXiv:2502.09294 [pdf, html, other]
Indeterminacy in Affective Computing: Considering Meaning and Context in Data Collection Practices
Bernd Dudzik, Tiffany Matej Hrkalovic, Chenxu Hao, Chirag Raman, Masha Tsfasman

Interface text:
arxiv.org/list/cs.AI/recent?skip=0&show=50
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Explore Pages: For the first five papers open as pdf's and extract title, top author, and summary
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek | Explore | Export
Advanced Options

[3:45](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=225s) [A spreadsheet shows extracted paper titles, authors, and summaries alongside the extraction assistant.](https://assets.ai.engineer/sites/aiecode2025/sha256/2d92285e308aa7f4a8d7814ae93751ff82b7b33b4668204517ba580514d7b7c0.webp)

Arxiv AI Paper Info [rtrvr.ai - 02/16/2025, 02:25 PM]

Parent Website | Source Website | paper_title | top_author | summary
https://arxiv.org/[illegible] | https://arxiv.org/[illegible] | CoT-Valve: Leng[illegible] | Xinyin Ma* | Chain-of-Thought significantly enhances a model's reasoning capability, but it also comes with a considerable increase in inference costs due to long c[illegible]
https://arxiv.org/[illegible] | https://arxiv.org/[illegible] | KIMAs: A Config[illegible] | Zitao Li | Knowledge-intensive conversations supported by large language models (LLMs) have become one of the most popular and helpful applications that c[illegible]
https://arxiv.org/[illegible] | https://arxiv.org/pdf/2502.09565
https://arxiv.org/[illegible] | https://arxiv.org/[illegible] | EMBODIEDBEN[illegible] | Rui Yang | Leveraging Multi-modal Large Language Models (MLLMs) to create embodied agents offers a promising avenue for tackling real-world tasks. While lar[illegible]
https://arxiv.org/[illegible] | https://arxiv.org/[illegible] | Dual Formulatio[illegible] | Navdeep Kumar | We study robust Markov decision processes (RMDPs) with non-rectangular uncertainty sets, which capture interdependencies across states unlike tra[illegible]

Interface text:
docs.google.com/spreadsheets/d/[illegible]/edit?gid=0#gid=0
File | Edit | View | Insert | Format | Data | Tools | Extensions | Help
Summarize this table
Sheet1
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Explore Pages: For the first five papers open as pdf's and extract title, top author, and summary
Explore Pages Response:
https://docs.google.com/spread...
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek | Explore | Export
Advanced Options

### Product page extraction

The demo starts extracting information from Amazon product tabs with an empty prompt. The speaker says the agent chooses fields to extract into Google Sheets for comparing the products.

[4:01](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=241s) [An earbuds product page is open while the assistant extracts information from tabs.](https://assets.ai.engineer/sites/aiecode2025/sha256/ecd6d46afe16f7ebd97f5221af6b38c2186d980a4092d88039b221c9ced77918.webp)

SAMSUNG Galaxy Buds 3 Pro AI True Wireless Bluetooth Earbuds, Noise Cancelling, Sound Optimization, Real-Time Interpreter, Redesigned Comfort Fit, Silver [US Version, Amazon Exclusive 2Yr Warranty]
Visit the SAMSUNG Store
4.1 stars · 982 ratings · Search this page
Amazon's Choice
10K+ bought in past month

Limited time deal
-20% $199.99
List Price: $249.99
FREE Returns
Available at a lower price from other sellers that may not offer free Prime shipping.

Color: Silver
$199.99; $249.99
$199.99; $249.99
Style: Buds 3 Pro
Pattern: Buds Only
Brand: SAMSUNG
Color: Silver
Ear Placement: In Ear
Form Factor: True Wireless
Noise Control: Active Noise Cancellation, Adaptive Noise Cancellation
About this item

9 VIDEOS
Roll over image to zoom in

Add Prime to get Fast, Free delivery
Delivery | Pickup
Buy new: $199.99
FREE Returns
FREE delivery Friday, February 21
Or Prime members get FREE delivery Tomorrow, February 17. Order within 9 hrs 4 mins.
Join Prime
Delivering to San Francisco 94115
Update location
In Stock
Quantity: 1
Add to cart
Buy Now
Ships from Amazon.com
Sold by Amazon.com
Returns 30-day refund/replacement
Payment Secure transaction
See more
Add a Protection Plan:
2-Year Protection Plan for $28.99

Interface text:
amazon.com/SAMSUNG-Cancelling-Optimization-Interpreter-Redesigned/dp/B0D9YZJ3V7?ref_=ast_sto_dp&th=1
Track price
Finish update
Add New · DEMO2 · git · Monitoring · All Bookmarks
Amazon · Delivering to San Franc... 94115 · Update location · Search Amazon · EN · Hello, sign in · Account & Lists · Returns & Orders · Cart
All · Medical Care · Best Sellers · Prime · Amazon Basics · New Releases · Music · Today's Deals · Groceries · Customer Service · Amazon Home · Pharmacy · Registry · Gift Cards · Books
Electronics · Cell Phones & Accessories · Cell Phones · Accessories · Cases · Wearable Technology · Deals
rtrvr.ai
Current Tab · Across Tabs · Sheets Flows
Extract Tabs:
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek · Explore · Export
Advanced Options

### Comparing earbuds

The demo runs an extraction across three Amazon product pages. The speaker says the agent selects useful fields and sends them to Google Sheets so the earbuds can be compared.

[4:14](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=254s) [A product page shows a comparison of earbuds while the extraction assistant runs.](https://assets.ai.engineer/sites/aiecode2025/sha256/d47c7f8586bd6f035cae2746d9b15b175ae210e7766d436b7757afa05272dc6d.webp)

Active Noise Cancellation – Bluetoo...
408
Limited time deal
-13% $199.00
List Price: $229.00
prime

Headphones Bass Stereo Ear Buds with Noise...
14,701
Amazon's Choice
$21.99
prime

Wireless Earbuds 60H Playback LED Power...
70,550
Limited time deal
-48% $25.99
List Price: $49.99
prime
Climate Pledge Friendly

Headphones in Ear buds, HiFi Stereo Wireless...
415
$29.99
prime
Save 45% with coupon

Wireless Earbuds with Wireless Charging Ca...
373,611
$24.99
prime
Save $5.00 with coupon
Climate Pledge Friendly

Headphones 45H Long Playtime, Wireless ...
225,236
$26.99
prime
Save $6.00 with coupon
Climate Pledge Friendly

Ear Buds 70H Playback Digital Display Wireles...
221
$22.99
prime
Save 30% with coupon

WARNING: California's Proposition 65

Top Brand: Google
Highly Rated
100K+ customers rate items from this brand highly
Trending
100K+ orders for this brand in past 3 months
Low Returns
Customers usually keep items from this brand

Compare with similar items
This Item
Google Pixel Buds Pro - Noise Canceling Earbuds - Up to 31 Hour Battery Life with Charging Case[2] - Bluetooth Headphone...
Add to cart
Price: -24% $152.99
List: $199.99

Recommendations
Google Pixel Buds Pro - Noise Canceling Earbuds - Up to 31 Hour Battery Life with Charging Case - Bluetooth Headphones - ...
Add to cart
Price: -45% $110.99
List: $199.99

Google Pixel Buds Pro 2 - Wireless Earbuds with Active Noise Cancellation - Bluetooth Headphones - Hazel
Add to cart
Price: -13% $199.00
List: $229.00

OnePlus Buds Pro 3, Wireless Bluetooth Earbuds, Noise Canceling Earbuds, Deep Bass, Immersive Sound, Up to 44 Hour...
Add to cart
Price: -11% $159.99
List: $179.99

Interface text:
amazon
Delivering to San Franc... 94115
Update location
Search Amazon
Hello, sign in
Account & Lists
Returns & Orders
Cart

Amazon.com: Apple AirPods Pro 2 Wireless Earbuds, Active Noise ...
amazon.com
Memory usage: 476 MB

Current Tab
Across Tabs
Sheets Flows
Extract Tabs:
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek
Explore
Export
Advanced Options

### Product comparison spreadsheet

A Google Sheet compares three earbuds across price, ratings, battery life and other fields. The selected image cell uses an Amazon image URL, illustrating the image extraction demo.

[4:34](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=274s) [A spreadsheet compares three earbuds using fields including price, rating, battery life, and connectivity.](https://assets.ai.engineer/sites/aiecode2025/sha256/45da628ae82e6c764fe691eda9362523433ed6e919384252f167501448bffe0d.webp)

Amazon Product Comparison [rtrvr.ai - 02/16/2025, 02:25 PM]

Source Website | image | product_name | brand | price | rating | number_of_ratin[illegible] | active_noise_ca[illegible] | battery_life | connectivity | color | special_featur[illegible]
https://www.ama[illegible] | [image] | Google Pixel Bu[illegible] | Google | 127.99 | 4.4 | 8416 | TRUE | Up to 11 hours o[illegible] | Wireless | Porcelain, Bay, [illegible] | Sweatproof, [illegible]
https://www.ama[illegible] | [image] | Apple AirPods P[illegible] | Apple | 199 | 4.6 | 19630 | TRUE | Up to 6 hours of [illegible] | Bluetooth 5.3 wi[illegible] | White | Dust, sweat, [illegible]
https://www.ama[illegible] | [image] | SAMSUNG Gala[illegible] | SAMSUNG | 199.99 | 4.1 | 982 | TRUE | Buds: 6H | Total: [illegible] | Wireless | Silver | Volume Contr[illegible]

Selected cell B2:
=IMAGE("https://m.media-amazon.com/images/I/[illegible]._AC_SX679_.jpg")

Interface text:
File  Edit  View  Insert  Format  Data  Tools  Extensions  Help
Share
rtrvr.ai
Current Tab  Across Tabs  Sheets Flows
Extract Tabs:
Extract Tabs Response:
https://docs.google.com/spread...
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek  Explore  Export
Advanced Options

### Extracting recent product reviews

With three Amazon tabs selected, the presenter asks the agent to switch from top reviews to the most recent and extract review details. Both visible pages still show Top reviews.

[4:46](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=286s) [An earbuds review page is shown beside an assistant with three product tabs selected and a review extraction instruction.](https://assets.ai.engineer/sites/aiecode2025/sha256/508fa3f0fc175e723bdd1bd83709208c12a8175aebe57f2234cc16463fff7aab.webp)

4 star 8%
3 star 3%
2 star 2%
1 star 6%
How customer reviews and ratings work

Select to learn more
Sound quality | Quality | Works well | Fit | Comfort | Hearing aid function | Battery life | Value for money

Reviews with images
See all photos

Review this product
Share your thoughts with other customers
Write a customer review

Top reviews
Top reviews from the United States

GusKlenke
★★★★★ These Are a Good Option
Reviewed in the United States on December 3, 2024
Verified Purchase

I've tried several hearing aid substitutes over the years, my past favorite being the now unavailable Bose Hearphones and costing $500. Rather than just a sound amplifier, these had several functions allowing you to filter and focus sounds in various ways and for various situations. They were pretty good.

But now Apple has added some great features to their AirPods 2—features that function like far more expensive hearing aids and at less than half the price of the Bose product. I've owned these for a month or so and am now just getting the hang of how to use them. If there is a downside to these earbuds it is the rather deep and confusing tech included in them (or, at least confusing for a grandfatherly type like me). Once they are Bluetooth connected to your iPhone (They connect very easily with the button on the front of the EarPod's charging case.) they should automatically connect every time you put them in your ears. You will then be able to access all the functions by tapping the “settings” app on your phone. You will find a heading near the top of the list just below the box with your name and iCloud options. Tapping on this will allow you to drill down through several pages of options. There are a lot of options there and it can be confusing, but experimenting with the various functions will help you figure out their purpose.
Read more
426 people found this helpful

Interface text:
amazon.com/dp/B0D1XD1ZV3/ref=fs_a_mdt2_us1
Top | About this item | Product information | Questions | Reviews
AI-generated from the text of customer reviews
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Search tabs...
new.express.adobe.com
rtrvr.ai pitch deck - Jan'25
amazon.com
Amazon.com: Google Pixel Buds Pro - Noise Can...
Amazon.com: Apple AirPods Pro 2 Wireless Earb...
Amazon.com: SAMSUNG Galaxy Buds 3 Pro AI Tr...
archive.ph
Several injured as car drives into people in Munich
Russia-Ukraine War: Trump Rushes Toward a De...
arxiv.org
Executable Code Actions Elicit Better LLM Agents
Artificial Intelligence
CoT-Valve: Length-Compressible Chain-of-Thou...
KIMAs: A Configurable Knowledge Integrated Mu...
MDCrow: Automating Molecular Dynamics Workfl...
EmbodiedBench: Comprehensive Benchmarking ...
Dual Formulation for Non-Rectangular L_p Robus...
Artificial Intelligence
bardeen.ai
How to scrape Zillow without code and gain an e...
bloomberg.com
Bloomberg - Business News, Stock Markets, Fina...
select by most reviews in the dropdown and extract the details of the most recent review
Ask | Research | Extract
Advanced Options

[5:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=321s) [An Amazon product page shows top reviews with several product tabs selected in the extension.](https://assets.ai.engineer/sites/aiecode2025/sha256/99d637824afde9bada23c42290ddd4d859a5d38979e2d370989186189e8ee784.webp)

2 star 4%
1 star 6%
How customer reviews and ratings work
Comfort | Fall out

Reviews with images
See all photos

Review this product
Share your thoughts with other customers
Write a customer review

Top reviews
Top reviews from the United States

MartialCyclist
★★★★★ Excellent and Sleek, but Not Perfect
Reviewed in the United States on December 29, 2022
Color: Charcoal | Verified Purchase

This is one of those situations where a 4.5 star rating would've been nice. And even though I point out a lot of ticky-tack issues, I'd by these again, no problem.

PROS:
- Excellent sound quality, especially for the price point and size
- Lightweight
- Premium design of both the buds and case
- Wireless charging on the case is a nice touch (especially since they kept wired charging with it)
- Excellent battery life
- Solid (if not perfect) noise cancellation
- Multi-device switching is awesome
- Single-bud use works great

CONS (and these are rather minor, but I'll cover them in detail):
Read more
282 people found this helpful
Helpful | Report
Anon

Interface text:
amazon.com/Google-Pixel-Buds-Pro-Headphones/dp/[illegible]
Top | About this item | Similar | Product information | Questions | Reviews
Amazon.com: Apple AirPods Pro 2 Wireless Earbuds, Active Noise ...
amazon.com
Memory usage: 244 MB
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Search tabs...
new.express.adobe.com
rtrvr.ai pitch deck - Jan'25
amazon.com
Amazon.com: Google Pixel Buds Pro - Noise Can...
Amazon.com: Apple AirPods Pro 2 Wireless Earb...
Amazon.com: SAMSUNG Galaxy Buds 3 Pro AI Tr...
archive.ph
Several injured as car drives into people in Munich
Russia-Ukraine War: Trump Rushes Toward a De...
arxiv.org
Executable Code Actions Elicit Better LLM Agents
Artificial Intelligence
CoT-Valve: Length-Compressible Chain-of-Thou...
KIMAs: A Configurable Knowledge Integrated Mu...
MDCrow: Automating Molecular Dynamics Workfl...
EmbodiedBench: Comprehensive Benchmarking ...
Dual Formulation for Non-Rectangular L_p Robus...
Artificial Intelligence
bardeen.ai
How to scrape Zillow without code and gain an e...
bloomberg.com
Bloomberg - Business News, Stock Markets, Fina...
Enter your prompt, for example:
- Ask: Use requirements.pdf tab to fill uptake form tab
- Research: Analyze the cloud offerings in these tabs
- Extract: Click more info, and extract address, price on each tab
Ask | Research | Extract
Advanced Options

### Reviews sorted by most recent

The agent switches the selected product tabs to Most recent. The visible Pixel Buds page now shows lilly’s February 13, 2025 review; the speaker says the recent reviews will be extracted.

[5:40](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=340s) [The product review page shows the most-recent sort and a recent customer review.](https://assets.ai.engineer/sites/aiecode2025/sha256/2eefe5515c9f4599ad91bb4a3f0aad1e2c1fd49ed1d269013369972ac4d44f7a.webp)

Customer reviews
4.4 out of 5
8,416 global ratings
5 star 70%
4 star 12%
3 star 7%
2 star 4%
1 star 6%
How customer reviews and ratings work

Customers say ai
Customers appreciate the headphones’ sound quality, battery life, and functionality. They find the sound decent, with good noise cancellation and controls closer to the cartilage of the ear. The battery life is great, with each earbud reporting the current level. However, some customers have issues with fallout. Opinions differ on fit, value for money, and comfort.
AI-generated from the text of customer reviews

Select to learn more
Sound quality | Battery life | Functionality | Build quality | Fit | Value for money | Comfort | Fall out

Reviews with images
See all photos

Review this product
Share your thoughts with other customers
Write a customer review

Most recent
From the United States

lilly
★★★★☆ Great!!
Reviewed in the United States on February 13, 2025
Color: Fog | Verified Purchase
I love the pixel buds. But my one complaint is the comfortability of them isn't that great. They somewhat hurt my ears and they come loose pretty easy. I don't use the sound quiet ability of the transparency feature because they make me feel sick. But the sound quality overall is good. I have dropped the buds a few times and they seem to hold up pretty good. They are super easy to pair with any device.
Helpful | Report

Interface text:
amazon
Delivering to San Franc... 94115
Update location
All | Search Amazon
EN
Hello, sign in
Account & Lists
Returns & Orders
Cart
Top | About this item | Similar | Product information | Questions | Reviews

rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Search tabs...
new.express.adobe.com
rtrvr.ai pitch deck - Jan'25
amazon.com
Amazon.com: Google Pixel Buds Pro - Noise Can...
Amazon.com: Apple AirPods Pro 2 Wireless Earb...
Amazon.com: SAMSUNG Galaxy Buds 3 Pro AI Tr...
archive.ph
Several injured as car drives into people in Munich
Russia-Ukraine War: Trump Rushes Toward a De...
arxiv.org
Executable Code Actions Elicit Better LLM Agents
Artificial Intelligence
CoT-Valve: Length-Compressible Chain-of-Thou...
KIMAs: A Configurable Knowledge Integrated Mu...
MDCrow: Automating Molecular Dynamics Workfl...
EmbodiedBench: Comprehensive Benchmarking ...
Dual Formulation for Non-Rectangular L_p Robus...
Artificial Intelligence
bardeen.ai
How to scrape Zillow without code and gain an e...
bloomberg.com
Bloomberg - Business News, Stock Markets, Fina...
Enter your prompt, for example:
Ask: Use requirements.pdf tab to fill [illegible] form tab
Research: Analyze the cloud offerings in these tabs
Extract: Click more info, and extract address, price on each tab
Ask | Research | Extract
Advanced Options

Browser address: amazon.com/Google-Pixel-Buds-Pro-Headphones/dp/[illegible]...
Add New | DEMO2 | git | Monitoring | All Bookmarks
Finish update

### Extracted reviews in a spreadsheet

The demo places the most recent reviews from three product pages into a spreadsheet, with each row recording a review’s title, author, rating, date and text.

[5:56](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=356s) [A spreadsheet contains extracted review titles, authors, ratings, dates and text from three product pages.](https://assets.ai.engineer/sites/aiecode2025/sha256/1aaec3f035b357628f21f35d757ae535926d9469b9932654103670bb47826d0d.webp)

Amazon Recent Review [rtrvr.ai - 02/16/2025, 02:26 PM]

Source Website | image | review_title | review_author | review_rating | review_date | review_text
https://www.amazon.com/Google-[illegible] | | Great!! | lilly | 4.0 out of 5 stars | Reviewed in the [illegible] | I love the pixel buds. But my one complaint is the comfortability of them isn't that great. They somewhat hurt my ears[illegible]
https://www.amazon.com/dp/B0D[illegible] | | Audio Issues | Mar | 1.0 out of 5 stars | February 15, 20[illegible] | AirPods does not work on video calls. It automatically disconnects every time...total waste of my money
https://www.amazon.com/SAMSU[illegible] | | Not too happy | Ro | 3.0 out of 5 stars | February 15, 20[illegible] | The design of the case and buds are very nice. Very easy to connect. But the fit was uncomfortable. But you might li[illegible]

Extract Tabs:
Extract Tabs Response:
https://docs.google.com/spread...

Extract Tabs: select by most reviews in the dropdown and extract the details of the most recent review

Extract Tabs Response:
https://docs.google.com/spread...

Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers

Interface text:
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
File | Edit | View | Insert | Format | Data | Tools | Extensions | Help
Summarize this table
Sheet1
Seek | Explore | Export
Advanced Options

### Summarizing selected design documents

The presenter selects two design documents and asks rtrvr.ai for key points and a summary. The response describes each document separately, including its components and intended users.

[6:17](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=377s) [A design document is open while two document tabs are selected and a summary request is entered in the extension.](https://assets.ai.engineer/sites/aiecode2025/sha256/c82860c5a60ef1f9ce54e3405d3ad864beb54b9b5014ab1ef390649933240f7d.webp)

## Design Document: Thinking LLM with Memory for AI Productivity Agents

**1. Introduction**

This document outlines the design and architecture of an AI agent system leveraging the power of Thinking Large Language Models (LLMs) combined with a robust memory system to enhance user productivity. We aim to create agents that can not only understand and generate text but also learn, remember, and reason about past interactions and user-specific data, leading to more personalized, efficient, and contextually aware assistance. This document is written from both an engineering and product perspective.

**2. Goals (Product)**

* **Enhanced Personalization:** Agents that adapt to individual user needs, preferences, and working styles.
* **Contextual Awareness:** Agents that remember past interactions, tasks, and information, eliminating the need for constant repetition.
* **Proactive Assistance:** Agents that anticipate user needs and offer relevant suggestions, information, or actions.
* **Improved Task Completion:** Agents that can effectively handle complex, multi-step tasks by leveraging memory and reasoning.
* **Seamless Integration:** Agents that integrate smoothly with existing productivity tools (e.g., email clients, calendars, project management software, note-taking apps).
* **Increased Efficiency:** Reduce time spent on routine tasks, information retrieval, and context switching.
* **Evolvability:** A system that can improve over time, with users opting into a feedback-and-improvement loop.
* **User Trust and Transparency:** Build trust by providing clear explanations for agent actions and offering control over data and privacy.

**3. Goals (Engineering)**

* **Scalable and Efficient Memory System:** Design a memory architecture that can handle large amounts of data with minimal latency.
* **Robust Context Management:** Develop mechanisms to efficiently retrieve and utilize relevant information from memory.
* **LLM-Memory Integration:** Seamlessly [illegible] memory system.
* **Modular Architecture:** Design a system that is easy to maintain, extend, and integrate with [illegible]

Interface text:
Design Document: Thinking LLM with Memory for AI Productivity Agents
File  Edit  View  Insert  Format  Tools  Extensions  Help
rtrvr.ai
Current Tab  Across Tabs  Sheets Flows
Search tabs...
adobe.com
Files | Adobe
new.express.adobe.com
rtrvr.ai pitch deck - Jan’25
rtrvr.ai pitch deck - Jan’25
chat.deepseek.com
DeepSeek - Into the Unknown
business.facebook.com
Message templates | WhatsApp Manager
developers.facebook.com
rtrvr.ai - WhatsApp - Meta for Developers
github.com
Updating default chatbox text, setting up agent h...
aistudio.google.com
AI Agent: Data Silo Solutions | Google AI Studio
docs.google.com
Ai Web Agent: Design Doc - Google Docs
Design Document: Thinking LLM with Memory fo...
Copy of Sheets Workflow Usecases - Google She...
rtrvr.ai: Buy Stocks! - Google Sheets
lu.ma
Events - Luma
extract key points and summary
Ask  Research  Extract
Advanced Options
meet.google.com is sharing your screen.  Stop sharing  Hide

[6:52](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=412s) [A design document beside an assistant response summarizing two selected documents.](https://assets.ai.engineer/sites/aiecode2025/sha256/086c6cea67f6e92968e68189ed45abbb44b0acade1daa91b1febf0d6d5c63b1b.webp)

system.
* **QA Engineers:** Develop and execute test plans to ensure the quality of the software.

This design document provides a comprehensive overview of the AI Web Agent. It is a living document that will be updated as the project evolves.

Ask Tabs: extract key points and summary

Ask Tabs Response: The first document (TabID 0) outlines the design for an AI Web Agent, a system that interacts with websites to automate tasks, extract information, and perform complex workflows. Key components include User Input, Task Planner (LLM), Action Executor, Browser Interface, Web Environment, State Evaluator, Reward Function, and a Feedback Loop (RL). It targets business professionals, researchers, software developers, individuals, and accessibility users. The technology stack includes Python, JavaScript, LLMs like GPT-4, computer vision, reinforcement learning libraries, headless browsers, and cloud platforms.

The second document (TabID 1) details the design of Thinking LLM with Memory for AI Productivity Agents, aiming to enhance user productivity through personalized, contextually aware assistance. Key components include a Thinking LLM Core (Foundation Model, Prompt Engineering, Action Generation), a Memory System (Short-Term and Long-Term Memory), User Interface/API, Agent Orchestrator, and a Security & Privacy Layer. It targets knowledge workers, executives, students, and individuals seeking personal productivity. Key technologies include LLMs, vector databases, graph databases, relational databases, caching, Python, JavaScript, and cloud platforms. The development [cut off]

Interface text:
Ai Web Agent: Design Doc
File  Edit  View  Insert  Format  Tools  Extensions  Help
rtrvr.ai
Current Tab  Across Tabs  Sheets Flows
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek  Explore  Export
Advanced Options
meet.google.com is sharing your screen.  Stop sharing  Hide

### Design-document summary continued

Retriever’s response summarizes each selected design document separately, then lists the memory-agent document’s components, intended users, technologies, and development roadmap.

[7:01](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=421s) [The assistant response scrolled to the second document’s details and development roadmap.](https://assets.ai.engineer/sites/aiecode2025/sha256/9b802aa3152f85baa73aff1dff5ba7b944cd769eb0b75fd3b63e40c327127ffb.webp)

system.
* **QA Engineers:** Develop and execute test plans to ensure the quality of the software.

This design document provides a comprehensive overview of the AI Web Agent. It is a living document that will be updated as the project evolves.

Web Agent, a system that interacts with websites to automate tasks, extract information, and perform complex workflows. Key components include User Input, Task Planner (LLM), Action Executor, Browser Interface, Web Environment, State Evaluator, Reward Function, and a Feedback Loop (RL). It targets business professionals, researchers, software developers, individuals, and accessibility users. The technology stack includes Python, JavaScript, LLMs like GPT-4, computer vision, reinforcement learning libraries, headless browsers, and cloud platforms.

The second document (TabID 1) details the design of Thinking LLM with Memory for AI Productivity Agents, aiming to enhance user productivity through personalized, contextually aware assistance. Key components include a Thinking LLM Core (Foundation Model, Prompt Engineering, Action Generation), a Memory System (Short-Term and Long-Term Memory), User Interface/API, Agent Orchestrator, and a Security & Privacy Layer. It targets knowledge workers, executives, students, and individuals seeking personal productivity. Key technologies include LLMs, vector databases, graph databases, relational databases, caching, Python, JavaScript, and cloud platforms. The development roadmap includes phases for MVP, enhanced memory, proactive assistance, and voice interface.

Interface text:
Ai Web Agent: Design Doc
File  Edit  View  Insert  Format  Tools  Extensions  Help
100%  Normal text  Arial  11
Share
rtrvr.ai
Current Tab  Across Tabs  Sheets Flows
Enter your prompt, for example:
- Seek: Find rtrvr.ai’s LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek  Explore  Export
Advanced Options
meet.google.com is sharing your screen.  Stop sharing  Hide

### Researching a company website

Axiom’s site is open while rtrvr.ai processes a request to analyze companies by strategy, features, pricing, and use cases—a market research example.

[7:16](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=436s) [An Axiom website page beside a company-analysis request and loading indicator.](https://assets.ai.engineer/sites/aiecode2025/sha256/84efbafdc78e2a54c2602458295a1bd6ffd4020280ba237af0fa9e80dc66f416.webp)

Video guides
Templates

Docs
Getting Started
How to automate web scraping guides
How to automate web forms
How to automate file uploads and downloads
How to send and receive data from your automations using APIs
Integrations

Automate the browser with a Macro Recorder alternative
See all

Web scraping
No-code data scraping from websites that require a login
Scrape data from multiple web pages
Scrape data from dynamic web pages
No-code data scraping from websites that require button clicks
See all

How to scrape ASIN data from Amazon
How to fill out forms online with a bot
Top tips for web automation in the Chrome Browser
See all

Other
Jobs
Affiliate program
Support
Contact
Desktop App

Find us online

Data, privacy and responsibility
All your bots live on your computer and process data in your web browser. We store the steps of your bot and data on how well it’s running, but we don’t store any of the results of your automations on our systems. Please use Axiom responsibly! Please read our terms of use and privacy policy for more information.

Backed by Y-Combinator

Research Tabs: Analyze these companies by strategy, features, pricing, and use cases

Interface text:
axiom.ai
Docs
Product
Pricing
Support
Blog
Install Chrome Extension

rtrvr.ai
Current Tab
Across Tabs
Sheets Flows
Enter your prompt, for example:
- Seek: Find rtrvr.ai’s LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek
Explore
Export
Advanced Options

meet.google.com is sharing your screen.
Stop sharing
Hide

### Researching another company

While rtrvr.ai processes a request to compare company strategy, features, pricing, and use cases, the browser shows Browse AI’s website.

[7:23](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=443s) [A Browse AI website page beside the same company-analysis request and loading indicator.](https://assets.ai.engineer/sites/aiecode2025/sha256/90eb9039cb1b90b4ace00e202086da731dc126e7d573dfeea17a5007acba422a.webp)

Which sites does Browse AI support?
How do you make sure my data is secure?

SOLUTIONS
Prebuilt robots
Popular use cases
Monitor websites for changes
Download data from any website
Turn any website into an API
Price monitoring
Integrations
Pricing
All features

RESOURCES
Help center
API documentation
Trust center
Blog
Careers
Contact us
Schedule a demo
About us
Media kit

PARTNERS
Affiliate program
Browse AI for startups

LEGAL
Privacy policy
Terms of service
Affiliate terms & conditions

SUBSCRIBE TO OUR NEWSLETTER
Receive the latest news, articles, and resources in your inbox monthly.
Enter your email
Subscribe
By subscribing, you agree to our Privacy Policy and provide consent to receive updates from Browse AI.

© 2025 Browse AI Inc.

Interface text:
browse.ai
Browse AI
Prebuilt robots
Features
Use cases
Resources
Pricing
Login
Get started for free
Help

rtrvr.ai
Current Tab
Across Tabs
Sheets Flows
Research Tabs: Analyze these companies by strategy, features, pricing, and use cases
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek
Explore
Export
Advanced Options

meet.google.com is sharing your screen.
Stop sharing
Hide

### Company research results

The demo starts with a company-analysis schema and fills the sheet with four companies. The presenter says rtrvr.ai visits pages such as pricing pages to find the requested information.

[7:30](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=450s) [A spreadsheet with company, strategy, features, pricing and use-case columns while research is loading.](https://assets.ai.engineer/sites/aiecode2025/sha256/15a7c83245f897a524ebb5ff28f4bb5c58c0633e0758bb1492bdbffbcbb9f1ac.webp)

AI Agent Company Analysis [rtrvr.ai - 02/16/2025, 02:28 PM]

company_name | strategy | key_features | pricing_model | use_cases

Research Tabs: Analyze these companies by strategy, features, pricing, and use cases

Interface text:
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Seek | Explore | Export
Advanced Options
Sheet1
meet.google.com is sharing your screen. Stop sharing | Hide

[8:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=501s) [A spreadsheet contains company analysis results, with some cell text truncated.](https://assets.ai.engineer/sites/aiecode2025/sha256/31eea2f8e5d48a1754d2265b1cf8360ea5b20f4ca81954e3fcd499ab280ec517.webp)

AI Agent Company Analysis [rtrvr.ai - 02/16/2025, 02:28 PM]

company_name | strategy | key_features | pricing_model | use_cases
rtrvr.ai | Automate web t[illegible] | AI-Powered Nav[illegible] | $0 Perfect for try[illegible] | Competitive Analysis: Open tabs for various competitor websites, including those requiring login. Use a query like “Analyze these companies[illegible]
Bardeen AI | Unlock superhur[illegible] | AI automation, w[illegible] | Free, Pro, Busin[illegible] | Automating lead discovery, enrichment and outreach sequences. Tracking account activity and nurturing target accounts. Preparing for custo[illegible]
Browse AI | The easiest way[illegible] | Data extraction,[illegible] | Free, Starter, Pr[illegible] | Scraping sales leads from online directories and marketplaces, Extracting and monitoring lists of products and their prices on e-commerce si[illegible]
axiom.ai | Save time - use[illegible] | Visual Web Scra[illegible] | Get started for F[illegible] | Data collection and analysis, Data extraction with GPT and data entry, Browser process automation

Selected cell B5: Save time - use browser bots to automate website actions and repetitive tasks on any website or web app.

Research Tabs: Analyze these companies by strategy, features, pricing, and use cases
Research Tabs Response:
https://docs.google.com/spread[illegible]

Interface text:
rtrvr.ai
Current Tab | Across Tabs | Sheets Flows
Seek | Explore | Export
Advanced Options
Sheet1
meet.google.com is sharing your screen. Stop sharing | Hide

### Stock research workflow

The presenter selects a sheet containing Chipotle and Starbucks websites for a stock research workflow, requesting P/E ratios, two years of revenue, and calculated revenue growth.

[8:44](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=524s) [A spreadsheet lists two companies beside a Sheets workflow configuration panel.](https://assets.ai.engineer/sites/aiecode2025/sha256/76cd9ff834120b34f3a40087722a5dc4b0747addd5a957fd558be19589af8f86.webp)

rtrvr.ai: Buy Stocks!

Companies
https://www.chipotle.com/
https://www.starbucks.com/

Sheets Flows
Select Previous Sheets Workflow
Step 1
Custom Workflow
rtrvr.ai: Buy Stocks!
Sheet2
Companies
First row has column headers
Delete tabs after execution
RegEx URL Substitution [Optional]:
Find: ^https://
Sub: https://arch[illegible]
Context Columns [Optional]:
None
Companies
Add Step
Clear All
Run

Interface text:
File Edit View Insert Format Data Tools Extensions Help
Current Tab
Across Tabs
Sheets Flows
Sheet1
Sheet2
meet.google.com is sharing your screen.
Stop sharing
Hide

### Stock research results

The sheet shows P/E ratios of 51 and 36.31, while the revenue columns are blank. The presenter says the agent corrected the supplied company URLs by finding Yahoo Finance.

[9:30](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=570s) [A spreadsheet shows two P/E ratios and columns for revenue and revenue growth.](https://assets.ai.engineer/sites/aiecode2025/sha256/44257e3a15edf0883c5a1a72951d054c3f47460c0edecf6773a14a3d62b87d40.webp)

Companies | pe_ratio | revenue_2024 | revenue_2023 | revenue_growth
https://www.chip[illegible] | 51
https://www.star[illegible] | 36.31

Selected cell formula: https://www.starbucks.com/

Interface text:
rtrvr.ai: Buy Stocks!
File  Edit  View  Insert  Format  Data  Tools  Extensions  Help
Current Tab  Across Tabs  Sheets Flows
Research Tabs: Analyze these companies by strategy, features, pricing, and use cases
[illegible] rtrvr.ai: Buy Stocks!
Clear Message
Clear All Before
Clear All
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek  Explore  Export
Advanced Options
Sheet1  Sheet2
meet.google.com is sharing your screen.  Stop sharing  Hide

### Stock results and assistant suggestions

The presenter says the agent corrected a supplied company URL and found the data elsewhere. The sheet shows two P/E ratios; the revenue columns are still empty.

[9:34](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=574s) [The stock research sheet remains visible beside assistant suggestion buttons.](https://assets.ai.engineer/sites/aiecode2025/sha256/8752b0dbf3e374ecc79a0ef7ecb6425150502980b80ae2129504aa05569ebec1.webp)

rtrvr.ai: Buy Stocks!

Companies | pe_ratio | revenue_2024 | revenue_2023 | revenue_growth
https://www.chip[illegible] | 51
https://www.star[illegible] | 36.31

Selected cell A3: https://www.starbucks.com/

Interface text:
File  Edit  View  Insert  Format  Data  Tools  Extensions  Help
Sheet1  Sheet2
rtrvr.ai
Current Tab  Across Tabs  Sheets Flows
Seek and follow rtrvr.ai on LinkedIn
Fill job applications using ask across tabs
Explore top Arxiv AI PDF papers
Do Deep Tab Research on rtrvr.ai
Export HackerNews Articles to Sheets
Auto Compare and Extract Products
Function Calling popup example
Use GraphBot to make in-chat graphs
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek  Explore  Export
Advanced Options
meet.google.com is sharing your screen.  Stop sharing  Hide

### Function calling configuration

The presenter shows configured functions, including WhatsApp messaging, and describes adding third-party tools by providing their details.

[9:39](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=579s) [A function-calling panel lists example functions beside Deep Tab Research documentation.](https://assets.ai.engineer/sites/aiecode2025/sha256/4be2a126f35ffb3b6051dfeca17308449881176c7a70748614ed87e42ceea833.webp)

Deep Tab Research with rtrvr.ai

Deep Tab Research is a groundbreaking feature in rtrvr.ai that allows you to conduct in-depth research using your own local tabs as starting points. Unlike traditional AI research tools such as Google Deep Research or Perplexity AI, rtrvr.ai can leverage login walled sites, private network sites, and local files, providing unparalleled flexibility and depth in your research process.

Core Concepts

Here are the key concepts for using Deep Tab Research:

• Local Tab Integration: Use your own open tabs, including those behind login walls or containing private information, as the basis for your research.
• Natural Language Queries: Provide research queries in natural language, such as "Analyze these companies by strategy, features, pricing, key differentiators".
• Active Research: rtrvr.ai not only extracts information from your selected tabs but also actively takes actions on them to further research and answer your query.
• Structured Output: Receive your research findings in a well-organized Google Sheet, with a table comparing your research topics against your desired fields.

Function Calling

Added Functions

graphBot: with provided context (tabs, sheets) and user prompt
slackMessage: Send a Slack message to configured channel with message parameter
testAlert: Creates a pop-up alert to display the message.
whatsAppMessage: send message to a whatsapp number

Shared Functions
Name  URL

Create Function
Function Name
Function Description

Interface text:
rtrvr.ai/docs/deep-tab-research
rtrvr.ai
Blog  Docs  Pricing  Trust & Safety  Install Extension  Request Demo
Current Tab  Across Tabs  Sheets Flows
meet.google.com is sharing your screen.  Stop sharing  Hide

[10:07](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=607s) [Research documentation beside a list of configured functions.](https://assets.ai.engineer/sites/aiecode2025/sha256/8ac308d13a4a7211952bc6ca647ac1cc7b6c656ececf1b2b4301b8d76dcaccb4.webp)

Deep Tab Research with rtrvr.ai

Deep Tab Research is a groundbreaking feature in rtrvr.ai that allows you to conduct in-depth research using your own local tabs as starting points. Unlike traditional AI research tools such as Google Deep Research or Perplexity AI, rtrvr.ai can leverage login walled sites, private network sites, and local files, providing unparalleled flexibility and depth in your research process.

Core Concepts

Here are the key concepts for using Deep Tab Research:

• Local Tab Integration: Use your own open tabs, including those behind login walls or containing private information, as the basis for your research.
• Natural Language Queries: Provide research queries in natural language, such as "Analyze these companies by strategy, features, pricing, key differentiators".
• Active Research: rtrvr.ai not only extracts information from your selected tabs but also actively takes actions on them to further research and answer your query.
• Structured Output: Receive your research findings in a well-organized Google Sheet, with a table comparing your research topics against your desired fields.

graphBot: with provided context (tabs, sheets) and user prompt
slackMessage: Send a Slack message to configured channel with message parameter
testAlert: Creates a pop-up alert to display the message.
whatsAppMessage: send message to a whatsapp number

Shared Functions
Name  URL

Create Function
Function Name
Function Description
Can LLM Auto-Call Function:
LLM and User  Only User

Interface text:
rtrvr.ai/docs/deep-tab-research
rtrvr.ai
Blog  Docs  Pricing  Trust & Safety  Install Extension  Request Demo
Current Tab  Across Tabs  Sheets Flows
meet.google.com is sharing your screen.  Stop sharing  Hide

### WhatsApp test message

The assistant reports that all messages were sent. The presenter opens WhatsApp and shows a 2:32 PM sample message on her test number.

[10:58](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=658s) [A WhatsApp conversation displays a sample message beside the demo assistant.](https://assets.ai.engineer/sites/aiecode2025/sha256/a08458915e3bc86afbc7023736a2c023e353250220db3aa517cfab55d927c15d.webp)

Meta. Thank you for taking the time to test with us.
WhatsApp Business Platform sample message
2:09 PM

Hello World
Welcome and congratulations!! This message demonstrates your ability to send a WhatsApp message notification from the Cloud API, hosted by Meta. Thank you for taking the time to test with us.
WhatsApp Business Platform sample message
2:25 PM

TODAY

Hello World
Welcome and congratulations!! This message demonstrates your ability to send a WhatsApp message notification from the Cloud API, hosted by Meta. Thank you for taking the time to test with us.
WhatsApp Business Platform sample message
1:36 PM

1 UNREAD MESSAGE

Hello World
Welcome and congratulations!! This message demonstrates your ability to send a WhatsApp message notification from the Cloud API, hosted by Meta. Thank you for taking the time to test with us.
WhatsApp Business Platform sample message
2:32 PM

Interface text:
web.whatsapp.com
rtrvr.ai
Current Tab
Across Tabs
Sheets Flows
Agentic Seek: send emssage to customer phone numbers here
Agentic Seek Response: All messages sent.
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek
Explore
Export
Advanced Options
Select Previous Task
Start or Select Recording Name
Show
Input
Output
Copy of Sheets Workflo...
Customers <> Whatsapp
Clear
Type a message
meet.google.com is sharing your screen.
Stop sharing
Hide
Bhavani Kalisetty

### Population graph demonstration

GraphBot turns a table of 2023 U.S. state population estimates into a bar chart. California leads, followed by a steep drop across the states.

[11:36](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=696s) [A browser extension displays a bar chart of estimated U.S. state populations beside a population table.](https://assets.ai.engineer/sites/aiecode2025/sha256/12e23523d0205e3a87374c366ae947fa2eae059e3099468c27086b834cff1820.webp)

list of U.S. states by population

[illegible] the largest population is California, which has some 39,000,000 residents. Second is Texas, with more than 30,000,000. Several states have less than 1,000,000 people, and Wyoming holds the title of the least-populated state, with fewer than 600,000. How do the other states rank? Find out in the table below, which lists the 50 U.S. states by the latest population estimate.

U.S. state | population: estimate | population: census
1. California | (2023 est.) 38,965,193 | (2020) 39,538,223
2. Texas | (2023 est.) 30,503,301 | (2020) 29,145,505
3. Florida | (2023 est.) 22,610,726 | (2020) 21,538,187
4. New York | (2023 est.) 19,571,216 | (2020) 20,201,249
5. Pennsylvania | (2023 est.) 12,961,683 | (2020) 13,002,700
6. Illinois | (2023 est.) 12,549,689 | (2020) 12,812,508
7. Ohio | (2023 est.) 11,785,935 | (2020) 11,799,448
8. Georgia | (2023 est.) 11,029,227 | (2020) 10,711,908
9. North Carolina | (2023 est.) 10,835,491 | (2020) 10,439,388
10. Michigan | (2023 est.) 10,037,261 | (2020) 10,077,331
11. New Jersey | (2023 est.) 9,290,841 | (2020) 9,288,994
12. Virginia | (2023 est.) 8,715,698 | (2020) 8,631,393
13. Washington | (2023 est.) 7,812,880 | (2020) 7,705,281
14. Arizona | (2023 est.) 7,431,344 | (2020) 7,151,502
15. Tennessee | (2023 est.) 7,126,489 | (2020) 6,910,840
16. Massachusetts | (2023 est.) 7,001,399 | (2020) 7,029,917
17. Indiana | (2023 est.) 6,862,199 | (2020) 6,785,528
18. Missouri | (2023 est.) 6,196,156 | [illegible]

GraphBot: @graphBot(prompt="please make a bar graph of the us states by population")

US State Populations (2023 Estimates)
Legend: 2023 Estimated Population (in millions)
Y-axis: Population (Millions)
Y-axis ticks: 0, 5, 10, 15, 20, 25, 30, 35, 40
X-axis labels: California, Illinois, New Jersey, Massachusetts, Colorado, Kentucky, Iowa, New Mexico, New Hampshire, South Dakota

Interface text:
britannica.com/topic/largest-U-S-state-by-population
Britannica
Search Britannica...
SUBSCRIBE
Login
Ask the Chatbot a Question
Table of Contents
Introduction
References & Edit History
Related Topics
Images
Read Next
Have Any U.S. Presidents Decided Not to Run For a Second Term?
What State Is Washington, D.C. In?
14 Questions About Government in the United States Answered
U.S. Death Toll During Major Events
26 Decade-Defining Events in U.S. History
Discover
Falling Far from the Tree: 7 Brilliant Ways Seeds and Fruits Are Dispersed
Who Really Invented Baseball?
Nikola Tesla's Weird Obsession with Pigeons
Was Jesse Owens Snubbed by Adolf Hitler at the Berlin Olympics?
See all related content
rtrvr.ai
Current Tab
Across Tabs
Sheets Flows
Enter your prompt, for example:
- Seek: Find rtrvr.ai's LinkedIn page
- Explore: For every pdf in directory, extract authors, summary, model
- Export: names, companies, titles, and intro email to speakers
Seek
Explore
Export
Advanced Options
meet.google.com is sharing your screen.
Stop sharing
Hide

### Redefining Web Data Extraction

The speaker positions Retriever’s text-based browser extension against cloud and vision-based agents, claiming lower cost, personalized results, access to paywalled sites, and processing across background tabs.

[12:05](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=725s) [A four-quadrant comparison of text- and vision-based extraction using cloud and user-controlled browsers.](https://assets.ai.engineer/sites/aiecode2025/sha256/f74bf4b7af1c876df4e2cf72fbc04435a6f2c7441ece2dd2e0947eef52bf4587.webp)

Redefining Web Data Extraction

Browser on-Cloud + Proxy

Text-based
1. Non-personalized results
2. No actions/just text extraction
3. Not fully agentic
Browserbase
Gumloop

Vision-based
1. More prone to hallucination
2. Highly expensive
3. Non-personalized results
Claude
Browse AI
OpenAI
bardeen

User-Controlled Browser + No Proxy

Text-based
1. Lower cost
2. High accuracy
3. Background & Multi tab processing
4. Supports paywalled & 1P sites
5. Personalized results

Vision-based
1. More prone to hallucination
2. Highly expensive
3. No background processing
4. No multi-tab support
MultiOn

Google

### Our mission

rtrvr.ai’s long-term goal is an AI-powered exchange where people collaborate across their own laptops to build datasets.

[16:30](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=990s) [The mission is to revolutionize data extraction with a transparent and efficient AI-powered exchange.](https://assets.ai.engineer/sites/aiecode2025/sha256/71017f0eb298517cb226103976f46329adec99551841475673846c8079a14ff8.webp)

Our mission

To revolutionize data extraction with a transparent and efficient AI-powered exchange.

### What's next?

The roadmap sets three stages over 6–18 months, moving from $500,000–$1M to $10M–$20M in ARR alongside customer targets and planned agent capabilities.

[16:47](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1007s) [Three-stage roadmap showing revenue, customer, and product milestones over 6 to 18 months.](https://assets.ai.engineer/sites/aiecode2025/sha256/37d59405c0f871be0923b83d9b72562b6460c0184cb088a8b9dcca6242bd77bd.webp)

What's next?

1
• $500,000 - $1M in ARR
• Acquire 250 paying SMB customers.
• Achieve an average MRR of $5K
• Launch Observable Agents
• Launch Custom Data Agents and Agent Evals
• Train MCTS Probability Model
6 -9 months

2
• $3M- $5M in ARR
• 50 mid-market clients. Avg. ACV $30K
• Continued growth in the SMB segment
• Support Browser on Cloud & API
• Increased investment in sales and marketing
9 - 12 months

3
• $10M- $20M in ARR
• Secure 10 enterprise clients. Avg ACV $75K
• Launch Federated Scraping Web Agent Marketplace
• Support Model Fine-tuning
12- 18 months

### Unlock the Full Potential of Your Browser

The speaker invites viewers to download the rtrvr.ai browser extension or scan the QR code to try it.

[17:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1041s) [Closing slide inviting viewers to try the browser extension, with a QR code and contact details.](https://assets.ai.engineer/sites/aiecode2025/sha256/8574b62862719e6ce19caf2000dd5f458fdb0e4f5d5a4aa703ca7f415c3377c3.webp)

rtrvr.ai: Unlock the Full Potential of Your Browser!

Contact: founders@rtrvr.ai
Support: support@rtrvr.ai

## Full transcript

[0:01](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1s) Hi, we're Arjun and Bhavani, and we're here to present to you rtrvr.ai. rtrvr.ai is a universal AI web agent that can do tasks autonomously on the web, extract structured data, and call APIs all for you with just natural language and with you as you browse, uh, like as a side panel with you, uh, with you as you

[0:19](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=19s) browse. So we believe this, uh, we could-- we'll dive deep into use cases as well as dive deep into how these kind of category of new universal AI web agents will transform people's workflows and be as transformative to the browser as its creation itself when Net-Netfl-Netscape, uh, came out.

[0:36](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=36s) So your bottle-- your browser is a bottleneck for most of your workflows right now. Even for people with full-time jobs, they're spending hours of their day manually copy and pasting from one website to another, w-uh, we-a website to like Google Sheets or to their CRM, or they even offshore the scraping to third parties that are expensive and

[0:55](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=55s) also unreliable. And, uh, yeah, and/or they set up RPA bots that just break whenever a website changes. And you also have silos where some data is only available on the website and some only on the APIs, leading to a hassle for users to combine this data across both, and the-- which leads to untapped potentials of what you

[1:16](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=76s) can do and what you can leverage this data for. So Retriever changes all this by being a, a Chrome extension that can d-- uh, that leverages being a AI web agent.

[1:26](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=86s) So you can open the side panel and do tasks acr-- uh, give tasks to do, uh, autonomously across pages, as well as extract structured data to Sheets. So let's dive into a use case.

[1:38](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=98s) So say you're on LinkedIn and you have our, uh, Chrome extension open. You can give a prompt just like, uh, for the Latent, uh, like find, uh, and follow the

[1:51](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=111s) Latent Space podcast page and, you know, if following, ignore, uh...

[2:02](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=122s) So just with this natural language prompt and clicking a button, our AI web agent will, uh, go to the task of, uh, interacting the, with the page to do this task.

[2:10](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=130s) So it can fill out the prompt f-- it can fill out the search field for Latent Space podcast. It can, uh, interact with all elements on the page. So it already would notice we're already following the page.

[2:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=141s) So, um, yeah. And you, uh, yeah, it, uh, even more-- We can do even more complex use cases than doing task autonomy. We can also have this built-in feature to extract data to Sheets.

[2:34](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=154s) So say you put this prompt, like for every article on the page, extract this data and click this export button. We would write all of this to Google Sheets for you.

[2:42](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=162s) And, uh, using, uh, our AI browser agent technology, you can also say like, "Do a task on this page and then extract the data." And, uh, like, you can have all of these combinations of tasks and extractions for you.

[2:56](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=176s) So it was able to extract all this data to Google Sheets, and we can say that ours, uh, our approach to solving this problem is very cost-effective, and it is probably cost less than a penny to do this page extraction.

[3:09](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=189s) And even more than doing actions on one page, we can do actions across tabs, across like Sheet, like you can give a Google Sheet column of URLs, and we can open those to interact with and extract from.

[3:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=201s) So we can lead to more complex use cases. So say you're on like, uh, this arXiv, uh, search, and you want for the first five PDFs, extract these, uh, fields.

[3:32](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=212s) We can break these down as sub-tasks to do and open them as new tabs and, uh, process these tabs simultaneously independent of each other and write this, uh, data to Sheets for you or interact with those pages.

[3:45](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=225s) Maybe they're, uh, LinkedIn job applications and you want to apply to them. So we are able to extract all of this data to Google Sheets and, um, you can do more complex use cases of you can select tabs to extract from.

[3:59](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=239s) So say you wanted to compare these Amazon product pages, and you could just leave the prompt empty itself, and our AI web agent will figure out what you-- fields you probably want to extract from.

[4:11](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=251s) And, um, from all these pages and extract those, uh, uh, pages to Google Sheets for you, extract those fields to Google Sheets, and so you can maybe effectively compare all three of these products without having to open them simultaneously.

[4:27](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=267s) And, uh, yeah, so we can even extract URLs on the page. So maybe you wanted to, uh, compare these, uh, products and you wanna see the images themselves, so we can extract the source image URLs, and you can compare them.

[4:40](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=280s) Even more than that, we can also, uh, do actions on the tabs themselves before extraction. So, so maybe you want instead of top reviews, uh, the most recent, so select, uh, by most recent, uh, reviews

[4:57](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=297s) in the dropdown and extract the details of the

[5:06](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=306s) most recent review. So these tabs are selected, and then I just do extract, and our AI web agent will do actions on all of these tabs simultaneously in the background and, uh, extract this, uh, most recent review that we wanted.

[5:22](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=322s) So they're all like on top reviews right now, and in the next action, it should go to most recent review. So yeah, it clicked most recent here and, uh, yeah.

[5:38](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=338s) So it clicked most recent across all of them, and it will give us the most recent review for us.

[5:44](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=344s) And even more than doing, uh, uh, actions just as basic extractions, we can also do, uh, um...

[5:55](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=355s) Wait, so... Yeah, so it was able to, uh, extract these most recent reviews, and Bhavani will dive into more advanced use cases of doing research and, uh, asking across these, uh, tabs you have.

[6:10](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=370s) Yeah. Um, if these use cases sounds, uh, exciting to you, let's unlock more potential of Retriever in your, like, day-to-day, like, at job tasks, right? Say for example, you have like bunch of design docs opened and you wanna send a quick summary to your, like, colleague at work.

[6:28](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=388s) All you have to do is select your design docs and then tell Retriever, "Okay, like, extract, uh, key points and summary." Right? Uh, and when you ask that, it goes through the docs that you like gi-- like selected and extracts the data that you like asked for.

[6:46](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=406s) So all unlocked productivity in like a one click of a button for you. Not just only Google Docs, you have like PDFs, you have Google Sheets. You can give any combination of anything that's on your browser and Retriever goes through them.

[6:58](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=418s) As you can see, it extracted, like, specifically what the first document talks about, and then what the second document about like design of like thinking with LLMs with memory as such.

[7:08](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=428s) Now, uh, not only with the existing, uh, documents that you have. Say you wanna do like a market research on a bunch of companies, uh, to like invest in also.

[7:19](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=439s) Like, here I'm asking it to go through a couple of agentic, uh, companies like Retriever, Bardeen, uh, Browse AI as such, and I'm asking it to extract certain things like strategy of the company, like features and pricing.

[7:32](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=452s) And what it does is first, based on what you asked, generates the schema, and then as you can see, we started with the homepage of all of these agent startups, and because we asked for like pricing, it navigated to like pricing page.

[7:44](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=464s) So Retriever is capable of going, uh, like basically like exploring the webpage or going deep into m-multiple pages to find the data that you asked for and extracts it right into the sheets for you.

[7:55](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=475s) And also to, uh... We are one of the very first in, uh, implementing this deep research feature compared to Operator also. Uh, and as you can see, the best part with our deep research is your ability to select all your, uh, documents that you have access to, not just on the browser.

[8:11](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=491s) And you can specifically give what all, uh, URLs or websites, uh, the agent can go in terms of like extracting this data for you. Uh, as you can see, successfully extracted all the information you asked for right into your Google Sheets for you.

[8:27](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=507s) Now, uh, like let's do this on like some steroids right? Now, like how about I want a market research, yes, but I want a market research about couple of stocks.

[8:37](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=517s) And then I-- you wanna like buy, make a purchase choice as such. So here all I have to do is select the right Google Sheet that has the information.

[8:45](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=525s) So for example, I have the sheet with couple of stocks lined up, and then I'm asking it like complex data. Like I'm asking it to give me like P/E ratio from Yahoo Finance and some revenue in the last two years, and I'm also asking it to create new data fields of computing the revenue growth, uh, based on

[9:02](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=542s) these numbers. So what it does, it's like first, um, uh, goes on, opens these websites and tries to like extract this data for you. Say, for example, right now I gave the websites as such.

[9:13](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=553s) Now, even though I gave the wrong URL, instead of giving Yahoo Finance, I just gave company's website, our agent figured out that it was not on the right URL and it went to like Yahoo, uh, and like kind of like tries to extract this data for you right into your sheets as such.

[9:30](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=570s) Now, so how cool is that? Like, even like there's like an error from the user, uh, Retriever was able to like correct this, uh, correct this for you. And one of the cool features we have is, uh, our function calling features.

[9:42](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=582s) Like, uh, lot of, uh, companies have like bunch of like third-party integrations with different tools, say Slack, uh, like say, say Discord also. One of the features we are very keen on implementing is like a dynamic function calling.

[9:56](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=596s) Like, instead of us implementing only like a set of, uh, connectors as such, you can call any API, any third-party tool out there, uh, by literally providing the information about the tool here and easily integrate.

[10:10](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=610s) Say, for example, I have WhatsApp integration right here to send messages to like WhatsApp numbers. Let's say you are a small business that you have like bunch of customers and like bunch of like customer phone numbers as such.

[10:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=621s) So I have two users here. And let's say I'm asking Retriever, uh, all I have to do is like select the sheets for you, share the sheets that has the customer numbers, and then I ask Retriever to like, uh, send message to customer phone numbers here.

[10:38](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=638s) So when I do that, uh, it identifies each of the customer number provided and invokes this, uh, WhatsApp tool, uh, essentially a third-party tool to send message to these users.

[10:48](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=648s) And as you can see, it says all messages sent successfully. Let, let's check it out. Let's try opening, uh, the WhatsApp.

[10:56](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=656s) So this is my test number, as you can see. Uh, we're demoing it at two thirty-two. So yeah, this is a two thirty-two like successful test message that we received.

[11:04](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=664s) So yeah, imagine the world of possibilities like, uh, automating your social communications across your Instagram, Facebook, WhatsApp, everything, all in like one click with Retriever right now. Uh, now let's do one more use case, uh, which is the graph generation.

[11:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=681s) So we have, uh, like a graph bot tool. Uh, we-- It's like almost like a mini agent within Retriever Agent Studio. What it does is, uh, let's say you have like a bunch of data as such, like, uh, and then you wanna generate like a data analysis graph on top of it.

[11:36](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=696s) You can leverage the graph bot in doing so. As you can see, it's like a dynamic graph generated on the fly for you. So we've realized when working with LLMs, they're not only inherently good with like data extraction as such, but are also capable of generating or representing this data in various formats of interest to the users,

[11:54](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=714s) and we leverage that capability to build such cool, uh, use case and present it to you. Now, like bringing it back home, now let's look at the overall agentic landscape, right?

[12:05](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=725s) Um, like let's look at, uh, some of like the Ope- OpenAI's Operator, like Anthropic Claude, and like obviously Retriever and so on. So most of the agents that are out there use like vision-based approach, like your Anthropic Claude or OpenAI, even like Google Mariner, who's a hybrid approach Also including vision approach.

[12:25](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=745s) As such, what do they do is they take screenshots of the pages and extract the data that you're asking for. And what's the inherent problem with this approach, right?

[12:33](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=753s) The vision-based models are more prone to hallucination compared to what Retriever is doing, which is text-based approach of leveraging the webpage stuff. And vision-based approach is also highly expensive, taking multiple screenshots for just like one single action to do or like for one single page scrolling as such.

[12:51](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=771s) And most of these companies such as these Browserbase or so on, they use browser on the cloud, unlike Retriever, which is an extension right inside your browser. So what are the problems with browser on the cloud?

[13:03](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=783s) Like number one, uh, non-personalized results, right? So it's like a generic page that's open on the browser. It's not like-- It mi- the content might be totally different from what you are seeing in your browser.

[13:14](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=794s) And also because, uh, to support browsers on the cloud, uh, these companies should implement a lot of proxies to funnel the n- uh, network or network request through the right IPs as such, which is way more expensive than the whole agentic setup as such.

[13:31](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=811s) Um, so with this, uh, text-based approach, as well as being like an extension in your browser, Retriever is capable of processing not only the active tabs, but also your background tabs, uh, or like multiple tabs at once.

[13:45](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=825s) And it can even go beyond your subscript- subscribed content. So with Retriever, you don't have to like share any of your passwords. We do not sh-store any of the passwords.

[13:54](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=834s) Whatever you are logged in and you are seeing, Retriever sees the same thing. Unlike on the cloud, um, either you have to like store or give the passwords, which are prone to like various security risks, or you cannot get behind paywalls or even like Cloudflare app website protections as such, right?

[14:12](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=852s) Now, like, A-Arjun will dive deeper into some of these and like, yeah, bring it back home.

[14:18](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=858s) Yeah. So, uh, just to recap what are the advantages for Retriever is that, uh...

[14:26](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=866s) Yeah, maybe I can, uh, share this. Uh, actually, it's fine. So, uh, to highlight the re- the advantage of Retriever, since we're using a text-based approach, the, the-- for writing output, there's, is, there's much less hallucination because the text is, uh,

[14:43](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=883s) the-- because the text is right there in context for the model. And what, uh, and because we are using this text-based approach, we can also take actions on multiple tabs because these background tabs don't get rendered.

[14:54](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=894s) So you've-- actually can't use a vision-based approach to take a- multiple actions in parallel. So when we can leverage this, uh, back- uh, background processing to do actions, uh, to do multi-tab context-aided actions, as well as just, uh, speed up performance by taking actions in parallel.

[15:12](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=912s) And in-- compared to our competitors of like trying to do consumer applications of booking flights or booking restaurant reservations, we're all in for productivity and automation use cases because we see that, uh, AI is perfect to automate these manual and repetitive tasks that is, uh, a burden to a lot of, uh, people.

[15:30](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=930s) And compared to everyone else, we are setting up a client-side Chrome extension that is, uh, not only cheaper infrastructure-wise, but it leverages us to use new sources of, uh, content to-- for the agent to work on.

[15:43](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=943s) So you can access the local walled sites or login walled sites, and, uh, it is much more secure than s-storing your passwords in a cloud-hosted browser. And compared to other providers of doing one single long horiz-- one t- uh, actions on a single tab with, uh, long horizon, uh, tasks, we distribute subtasks as new tabs to take

[16:04](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=964s) actions on. So our failure rate is much less than these, uh, uh, competitors. So we don't, don't have to deal with, uh, exponential failure rates. And our approach to third-party integrations is that we, the user can define and, uh, set up function calling that-- function calls that they can share.

[16:21](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=981s) And this is much more, uh, extensible and scalable than setting up custom third-party integrations that these, uh, third parties-- uh, that these competitors are doing.

[16:31](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=991s) Yeah, so this is our mission is to basically revolutionize data extraction with a transparent and efficient AI-powered exchange. This is our long-term goal of being a-- allowing people to collaborate across the-- their own local laptops to collaboratively construct datasets and, uh, set up, uh, uh, like very cost-efficient and cheap datasets.

[16:54](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1014s) So for example, it could be like all the, uh, local government events happening in the SF Bay Area that would involve like extracting data from hundreds, thousands of websites and this is not something that is feasible right now.

[17:06](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1026s) It's like not cost-effective to do. But if people can use it, leverage our extension to volunteer and collaboratively construct datasets, we believe that's an exciting new future and use case that is, uh, that is just on the horizon.

[17:22](https://www.youtube.com/watch?v=kjSGc7uwDo8&t=1042s) Thank you, and, uh, please feel free to go to rtrvr.ai to download the extension. That's Retriever without any vowels. Or you can scan this QR code and try it out and hopefully you can get a glimpse into what the future of, uh, AI agents in the browser will be.
