Company knowledge and AI agent memory
CentralAgent AI
CentralAgent AI builds shared company memory for engineering teams and their AI agents. It captures meetings and agent sessions alongside documents, code changes and production logs, preserving context such as why a deployment rule exists or which fix resolved an incident. Employees query that memory through CentralView, while coding agents including Claude Code, Cursor and Codex access it through a Model Context Protocol (MCP) endpoint. The aim is to make knowledge held by experienced colleagues available to new hires and fresh agent sessions.
Its defining mechanism is the conversion of conversation into structured, persistent memory. Records are dated and attributed, and a task loads the relevant portion rather than the entire company history. Superseded records are retired while their history remains available. CentralAgent combines three representations: a knowledge graph for retrieval with citations, a language-model-maintained wiki and a customer-specific model that is continually fine-tuned. Incoming questions receive answers from all three, which the system compares before applying correctness and personally identifying information checks.
CentralAgent deploys in the customer’s private cloud, with controls over the resources it can access. Teams can supply model API credentials or use models hosted within their cloud. Beyond capturing existing knowledge, its inference workflow proposes potential failure points and conflicting assumptions for senior engineers to confirm, reject or correct. Those judgments become citable records, adding the engineers’ reasoning to the memory available for later work.
1 talk
Newest firstLoading more talks and speakers…
1 speaker at AIE
Affiliations reflect their AIE event listings and recorded appearances, not necessarily current employment.
