# ai-memory

> Long-term memory for AI coding agents. Open source (MIT), written in Rust, by Fabio Akita. Quit one coding agent mid-task, start another in the same directory, and continue without re-explaining the project.

ai-memory is a single self-contained binary that runs a small MCP/HTTP server. Lifecycle hooks in the coding agent capture prompts, tool calls and session boundaries, sanitized before storage. Sessions are consolidated into a git-backed wiki of plain markdown files, which is the source of truth; SQLite holds a derived, rebuildable index (FTS5, entities, links, optional vectors). Capture, search and handoffs work with zero LLM calls and no API key. One server can be shared by several machines and by a team. The wiki is natively an Open Knowledge Format (OKF v0.2) bundle. Latest release: v2.4.0.

## Product

- [How it works](https://aimemory.io/#how): Capture, consolidate, recall, hand off
- [How memory ages](https://aimemory.io/aging/): Decay, compaction and the dream pass
- [Architecture](https://aimemory.io/architecture/): One binary, markdown files, a derived index
- [Security](https://aimemory.io/security/): What is stored, what leaves, how to lock it down
- [Research and rationale](https://aimemory.io/research/): Karpathy’s wiki, OKF, and the benchmarks

## Solutions

- [For individuals](https://aimemory.io/solutions/individuals/): Switch agents and machines, keep the context
- [For teams](https://aimemory.io/solutions/teams/): One shared memory, faster onboarding
- [Migrating from another tool](https://aimemory.io/compare/): Mem0, Zep, cognee, OpenViking and others

## Integrations

- [Agents and editors](https://aimemory.io/integrations/#agents): Claude Code, Codex, Cursor, Gemini CLI and 20 more
- [Operating systems](https://aimemory.io/integrations/#platforms): Linux, macOS, Windows
- [LLM and embedding providers](https://aimemory.io/integrations/#providers): Optional. Everything works without one

## Developers

- [Quick setup](https://aimemory.io/install/): Commands for one machine or a whole team
- [Advanced setup](https://aimemory.io/advanced/): Homelab, TLS, SSO, air-gapped, backups
- [Contribute](https://aimemory.io/contribute/): Where issues, PRs and new harnesses go
- [Downloads and changelog](https://aimemory.io/download/): Latest release for every platform

## Languages

- [English](https://aimemory.io/)
- [Português (Brasil)](https://aimemory.io/pt-br/)
- [Español](https://aimemory.io/es/)
- [עברית](https://aimemory.io/he/)
- [日本語](https://aimemory.io/ja/)
- [한국어](https://aimemory.io/ko/)

## Source and documentation

- [GitHub repository](https://github.com/akitaonrails/ai-memory): source, issues, pull requests
- [README](https://github.com/akitaonrails/ai-memory#readme): overview and quick start
- [Installation cookbook](https://github.com/akitaonrails/ai-memory/blob/main/docs/install.md): every agent and platform
- [Architecture](https://github.com/akitaonrails/ai-memory/blob/main/docs/ARCHITECTURE.md): data flow, crates, invariants, schema
- [How ai-memory compares](https://github.com/akitaonrails/ai-memory/blob/main/docs/comparison.md): a fair rundown against other memory tools
- [Benchmarks](https://github.com/akitaonrails/ai-memory/blob/main/docs/benchmarks/README.md): published LongMemEval-S retrieval numbers, reproducible
- [Security model](https://github.com/akitaonrails/ai-memory/blob/main/docs/security.md) and [data handling](https://github.com/akitaonrails/ai-memory/blob/main/DATA_HANDLING.md)
- [Changelog](https://aimemory.io/download/#changelog) and [releases](https://github.com/akitaonrails/ai-memory/releases)