Summary

akitaonrails/ai-memory addresses AI coding agent context loss via a persistent, shared wiki for long-term memory. It captures prompts, tool calls, and decisions, enabling seamless handoffs between agents/sessions. Users resume tasks without re-explaining context, enhancing productivity and continuity.

How It Works

Lifecycle hooks capture prompts, tool calls, and session boundaries, compiling them into plain markdown pages within a Git repository, forming a versioned "LLM wiki." This eliminates complex vector databases or manual context management. A Rust server processes observations, indexes them in SQLite (FTS5, optional embeddings), and serves retrieval, enabling cross-agent continuity and time-travelable project history.

Quick Start & Requirements

Installation via Arch Linux AUR (ai-memory-bin, ai-memory) or Docker. Docker quick-start needs a CLI wrapper and server container (loopback default). Prerequisites: Docker, compatible agent CLI (Claude Code, Codex, Cursor, Gemini CLI). Optional LLM/embedding API keys for advanced features. Setup guides: docs/install.md, docs/deploy.md.

Highlighted Details

• Zero-Friction Capture: Automatic logging of prompts, tool calls, and session boundaries via lifecycle hooks.
• Cross-Agent Handoffs: Resume tasks with different agents in the same project, prepending a "where you left off" summary.
• Per-Project Isolation: UUID-based isolation, configurable via .ai-memory.toml for monorepos/multi-client setups.
• Git-Versioned Markdown Wiki: History stored as markdown in Git, allowing time-travel and direct inspection (e.g., Obsidian).
• Web UI: Read-only HTML interface for browsing, FTS5 search, and markdown rendering.
• Multi-Agent/Machine Support: Server runs locally or networked, accessible by multiple clients, with optional bearer token auth.
• LLM Integration: Optional LLM providers enhance consolidation; functional Zero-LLM mode available.

Maintenance & Community

Built collaboratively with Claude Code. Specifics on maintainers, community channels, or sponsorship are not detailed in the provided README.

Licensing & Compatibility

MIT license, permissive for commercial use and integration. Supports numerous LLM providers and agent CLIs.

Limitations & Caveats

Native Windows support is "Experimental." Users advised to use WSL2 or PowerShell wrapper. No other significant limitations highlighted.