Context repository
A context repository is the self-maintaining, versioned store of knowledge that AI agents read from.
Category
Product feature
See all terms
What is a context repository?
A context repository is a workspace's central store of knowledge, whether that workspace belongs to one person or an entire company, organized as files and folders and connected through a context graph, rather than scattered across separate tools and conversations. It holds normalized information about entities like customers, employees, products, and projects, built from whatever data sources a workspace connects, and every AI agent or Client reads from that same structured store instead of each tool holding its own partial copy.
Why does a context repository exist?
Without a shared repository, every team and every AI tool builds its own version of the same company knowledge, each drifting out of sync with the others and with reality. A context repository exists so there's one place that's authoritative, current, and reachable by every agent, closing the gap between what a company actually knows and what any single tool happens to have on hand.
How does a context repository work?
A context repository stores knowledge as files with version history, so every change is tracked. Updates can be drafted, reviewed, and reconciled automatically as new information arrives from connected sources, and continuous updates keep files current without anyone re-running an import by hand. Access controls apply per user and per group, and every Client, an API key or an MCP client, reads through the same permissions rather than a separate, unmanaged path. Retrieval logs record what was read and by whom, giving the repository an auditable trail on top of being current.