OKF Agent Memory

OKF Agent Memory gives AI coding agents persistent project knowledge across sessions. It's for developers who want agents to retain architectural decisions and project constraints without loading the entire knowledge base into every prompt.

• Rules and retrieved knowledge: Keep compact operating instructions in the agent's starting context while letting it retrieve linked knowledge files when a task needs them.
• Git-based memory: Store knowledge as version-controlled plain text so you can inspect edits, review changes and trace their history alongside code.
• Local search and agent connections: Search with BM25 without an external database or embedding API, and connect the embedded MCP server to Claude Code, Cursor or Codex.
• Sources and validation: Record where knowledge came from, distinguish generated content from verified facts, track status and staleness, and validate the knowledge graph.
• Scoped project constraints: Separate project knowledge, external packages, personal preferences and machine policies, with project decisions taking priority when concept IDs overlap; link decisions to source files so agents can retrieve relevant constraints before editing.

The tool is open source under the MIT license and uses the Open Knowledge Format. It's written in Go and provides a standalone command-line tool. Storage and retrieval run locally without network calls. Its search-before-write convention directs agents to check existing memory before adding concepts, helping reduce duplicate entries.