akitaonrails/ai-memory is a Rust server that gathers the long-term memory of agent coding CLIs in one place. According to the provided information, the latest release is v3.1.0, but what changed in this version is not given. The repository is MIT licensed and has 9609 GitHub stars.
This server manages memory shared by more than 20 agent runtimes. The source of truth for memory is not a vector database but a markdown wiki backed up by git. So you can find files with grep, edit them directly or restore them, and the database is only a derived index that can be rebuilt. The README gives the example of continuing work done in Claude Code with Codex in the same directory, and explains that remaining tasks, failed attempts, and open questions are put into shared memory. Handoff information has a type assigned and is processed only once.
Tracing, search, and handoff on the default path work without model calls, apply a stepwise decay rate to memory, and make frequently used notes age more slowly. The repository includes 11 crates plus hooks, evals, and tests. However, the provided information contains no figures for judging search quality or performance. So the design that lets you manage files directly is confirmed, but it is hard to judge from this material alone how useful the memory actually is in a real environment.