A consistent, versioned knowledge base of an organization's data, built from its many sources and read by people and agents.
The data of a large organization is scattered across many systems: tickets, documentation, code, monitoring, meetings. It keeps changing, parts of it are stale or contradict each other, and it is hard to stitch together. AKMON builds and maintains one knowledge graph of it, the operational knowledge graph (OKG), so that people, agents and software can query it.
sources ─▶ incremental ingestion ─▶ consistency checks ─▶ versioned graph ─▶ server ─▶ people and agents
AKMON runs on PostgreSQL, on your own machines. Agents connect to it over the
Model Context Protocol (MCP); people use the okg command.
The Quickstart takes you from nothing to a graph
of your own repository and documents, with your coding agent connected. You
need Linux or macOS, Git, uv, and Docker or
podman.
- Get started and the AKMON model
- Connect your data
- Connect agents
- Run it as a service
- Reference
- Develop AKMON
Start with Develop AKMON and CONTRIBUTING.md. For changes to the import, publication, or query path, also read the code architecture.
AKMON is released under the MIT License; see LICENSE.