akitaonrails/ai-memory
A shared memory server so switching AI coding agents doesn't wipe out your project context
ai-memory is a Rust tool that lets you quit Claude Code mid-task and continue the same work in a different agent like OpenAI Codex, in the same directory, without re-explaining the architecture, failed approaches, or open questions. It hooks into each agent's session start/end lifecycle to capture sanitized observations, then compiles them into a summary page that gets handed off to whichever agent starts next. Everything is stored as plain markdown in a git repo rather than a vector database, so it's grep-able, viewable in Obsidian, and backup-able with rsync.
What it does
- It installs MCP configuration and lifecycle hooks into agents such as Claude Code, Codex, Cursor, Gemini CLI, and Devin CLI, capturing sanitized prompt and tool-use observations as each session runs.
- When a session ends, those observations are compiled into a coherent summary wiki page, and the next session automatically receives a 'where you left off' handoff block.
- Data is isolated per project under UUID-keyed paths, and search combines FTS5 full-text search with entity (noun) matching and graph-neighbor lookup, with optional LLM summarization and vector search.
- It's installed via CLI commands like install-mcp and install-hooks, and the server can run locally or on a home server so multiple machines and agents share the same memory store.
- The system works without any LLM using rule-based summarization and search alone, or you can plug in providers like Anthropic, OpenAI, or Gemini for richer consolidation.
Why it matters
Developers who bounce between multiple AI coding tools currently pay a real cost re-explaining project background, failed attempts, and open questions every time they switch — this tool preserves that context in a vendor-neutral form, cutting the switching cost. Its markdown-plus-git storage also means it slots into existing workflows without requiring new infrastructure.
Terms in this repo
- MCP · a standardized protocol that lets AI agents connect to external tools and data
- lifecycle hooks · connection points that auto-trigger at specific moments in an agent's run, like session start or end
- FTS5 · SQLite's built-in full-text search engine
- handoff · a summary passed from one session to the next describing where work was left off
- RRF · a method for merging rankings from multiple search results into one combined order
Repository description (English)
Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors
Open on GitHubTrending repos
- cathrynlavery/diagram-designA skill that makes AI coding tools draw magazine-quality diagrams instead of generic rounded boxes
- public-apis/public-apisA giant crowd-curated directory of free APIs for developers
- semantica-agi/semanticaAn open-source graph infrastructure that lets AI agents show their work, not just their answers
- cactus-compute/needleA 14MB AI model small enough to run tool-calling on a phone or watch, without internet
- unslothai/unslothA desktop app that lets you run and train AI models on your own computer, no coding required
- macro-inc/macroAn all-in-one workspace where email, chat, docs, tasks, and CRM are cross-linked and share one AI memory
- harry0703/MoneyPrinterTurboAn open-source tool that turns a single topic or keyword into a finished short video, complete with script, footage, subtitles, and music
- basecamp/omarchyA ready-made, opinionated Linux setup built by DHH
Latest from METAL LAB
- Grok leaks conversations, location data via commands hidden in ciphertext
- Cursor lets agents handle long-running tasks with new "/goal" command
- NVIDIA releases Switchyard, an LLM routing proxy for coding agents
- Caveman Cuts Claude Code Token Usage by 33% Using Caveman-Speak
- Apple Keeps 2027 Target for Camera AirPods, Qwen3.8-27B Makes 80 Tool Calls Locally