ai-memory is an open-source memory layer that sits outside your coding agents. Lifecycle hooks and MCP registration record what happens in a session (prompts, tool calls, session boundaries), sanitize it at a privacy boundary and consolidate it into readable markdown pages in a per-project wiki. The next session, in any supported agent and on any machine, gets a bounded brief and can search earlier work with full-text, entity and link search, optionally fused with vector search.
Its headline feature is cross-agent handoff. Quit Claude Code mid-task, open Codex in the same directory, and the next agent receives a handoff describing where you stopped, what failed and what is still open; handoffs are a typed protocol with an owner and are claimed exactly once. Starting agents through ai-memory run can also pass a temporary local projection of the previous agent's visible session history. A small cross-project profile carries habits such as your package manager and test layout into new repositories.
The source of truth is a git-backed wiki of ordinary .md files that you can grep, open in Obsidian or edit by hand; the database is a derived index that can be rebuilt. Memory ages on a tunable schedule without an LLM, and optional compaction, deduplication and an opt-in LLM "dream" pass that merges cold notes are off by default. For teams, one server provides shared per-project knowledge with multi-user auth, per-person attribution and an audit log, which the README says are not a paid tier.
ai-memory ships as one self-contained Rust binary for Linux and macOS (Windows via WSL2; native Windows is experimental), with install scripts that verify SHA-256 checksums, a Claude Code plugin, AUR/RPM packages, a macOS app and a Docker image. It is free and MIT-licensed; you run the server yourself, and optional cloud model or embedding providers bill under your own accounts. A default install listens only on 127.0.0.1 and, per the project's security page, has no analytics, crash reporting or usage tracking.
