OKF Agent Memory
Git-native persistent project memory for agents based on Open Knowledge Format: a single Go binary, an embedded MCP server and BM25 search
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Reads and writes project memory files under git
- Runs a local MCP server process
Install
Manual install
git clone https://github.com/okf-memory/okf-agent-memory.git
cd okf-agent-memory
make build
./bin/okf mcp knowledgeBuilds the bin/okf binary and runs the embedded MCP over stdio. A Homebrew formula is available.
This is third-party code. Review the repository files before installing.
What it does
OKF Agent Memory keeps project knowledge in the git-native Open Knowledge Format instead of bloating CLAUDE.md or a vector store. A dual-memory architecture splits normative working memory that the agent always holds from semantic domain memory that costs zero tokens at baseline and is queried on demand through okf_search and okf_show. The tool is built in Go as a single dependency-free binary made for high-frequency agent tool calls. It can validate and search knowledge, create and update entries, scaffold the memory structure into a project and run an embedded MCP server over stdio.
Who it is for. For developers who want fast, readable, git-stored project memory with no external database.
Good fit when
- You want project memory in git rather than a vector store
- CLAUDE.md or AGENTS.md have bloated and the agent drifts
- You need fast knowledge search right inside the agent loop
Not a fit when
- You need embedding-based semantic search rather than BM25 over markdown
- You are not ready to build the binary from Go source
Example request
Record the auth-flow decision into OKF memory and find related entries on the architecture layerLimitations
The tool is built from Go source; a package in standard managers may not exist, though a Homebrew formula is provided. Memory is markdown in OKF v0.2 format, and search is BM25 rather than embeddings. Optional Hub sync needs a token and a key.
How to disable. Remove the MCP entry from the agent config and delete the built binary; the project's knowledge/ folder can stay or be removed manually.
MCP
- Transport
- stdio
- Authentication
- not required
| Environment variables | |
|---|---|
| OKF_HUB_TOKEN secret | Token for optional memory sync through the Hub. |
Security check
- Reads and writes project memory files under git
- Runs a local MCP server process
README in short
The README describes OKF Agent Memory as git-native persistent project memory based on Open Knowledge Format v0.2. A dual-memory architecture splits normative working memory from semantic domain memory that costs zero tokens at baseline and is queried through okf_search and okf_show. The tool is built in Go as one dependency-free binary, fast in the agent loop, with provenance and trust tiers. It validates and searches knowledge, creates and updates entries, scaffolds the structure into a project and runs an embedded MCP server over stdio.
FAQ
Where is memory stored?
In the project's knowledge/ folder in OKF v0.2 format, plain markdown under git that is readable without the tool.
Is this semantic search?
Search is BM25 over the knowledge bundle, fast and without embeddings, honoring entry provenance and trust tiers.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything