OKF Agent Memory: agent memory as plain Markdown files
OKF Agent Memory Every agent session starts from zero, and the decisions from last week are lost unless someone wrote them down in the right place. OKF Agent Memory fixes that with plain files your team can already review. It implements Google's Open Knowledge Format v0.2 , so memory lives as Markdown with YAML frontmatter in a git tracked knowledge folder, and every entry records whether an…
Every time an agent session begins, it starts from scratch and discards any knowledge gained from the previous week unless someone documents those decisions in a designated location. OKF Agent Memory solves this issue by utilizing plain Markdown files that your team can easily review. It adheres to Google's Open Knowledge Format v0.2, with memory residing as Markdown files containing YAML front matter in a git-tracked knowledge folder.
Each entry also specifies whether an agent created it or if a human verified it. A small push layer restricts the active codex to 400 tokens, while the complete knowledge bundle is retrieved as needed, achieving sub-300 microsecond BM25 search and served over MCP to agents. Version 0.8.0 was released on November 5th, 2026, and the repository currently boasts 866 stars under the MIT license.
What sets OKF Agent Memory apart is its ability to present memory as straightforward Markdown files inside a git-tracked folder, similar to code that can be reviewed. Provenance is recorded for every entry, indicating whether the agent authored it or if a human confirmed it. The system uses a 400-token working codex, with the entirety of the knowledge bundle being fetched on request.
BM25 search boasts sub-300 microsecond response times and is delivered via MCP to your agents. By focusing on team decisions rather than chat logs and verifying an entry before the agent treats it as fact, OKF Agent Memory offers a superior alternative to memory locked inside a service.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
