akitaonrails/ai-memory

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

What it solves

AI coding agents typically lose context when a session ends, forcing users to re-explain architecture, failed attempts, and open questions when starting a new session or switching between different AI agents (e.g., moving from Claude Code to Codex).

How it works

ai-memory creates a shared, persistent wiki of project-specific knowledge compiled from sanitized lifecycle observations (prompts, tool use, and session boundaries). It uses a Model Context Protocol (MCP) server to capture these events via hooks and summarize them into plain Markdown files stored in a git repository. This allows the memory to be grep-able, version-controlled, and accessible via a read-only HTML UI. It supports a wide array of AI agents and CLIs, providing "handoffs" that tell the next agent where the previous one left off.

Who it’s for

Developers using multiple AI coding agents or CLIs (such as Claude Code, Codex, Cursor, Devin, and others) who want to maintain a continuous, cross-agent workstream without losing historical context.

Highlights

  • Cross-Agent Continuity: Seamlessly transition between different AI agents while preserving context via handoffs and managed workstreams.
  • Zero-Friction Capture: Uses lifecycle hooks to automatically capture and sanitize prompt and tool events without manual note-taking.
  • Markdown-Based Wiki: Stores memory as plain Markdown in a git repo, avoiding the case for a vector database and allowing for easy backup and versioning.
  • Broad Agent Support: Compatible with a vast range of tools including Claude Code, Codex, Command Code, Devin CLI, Cursor, Gemini CLI, and many others.
  • Advanced Recall: Combines FTS5, entity-match RRF, and authority-aware retrieval to prioritize maintained rules and decisions over episodic session evidence.
  • Managed Workstreams: An optional launcher (ai-memory run) that provides native session resume and a portable visible-event ledger.
  • Opt-in LLM Integration: Can operate in a Zero-LLM mode using lexical search, with optional LLM providers for consolidation and auto-improvement.

Related

  • Project
    JamesShi96/project-butler

    A persistent memory system for AI coding assistants that uses a structured Markdown-based memory stack to maintain project context, rules, and progress across sessions.

  • Project
    rohitg00/agentmemory

    A persistent memory server for AI coding agents that automatically captures and retrieves project context across sessions using hybrid search.

  • Project
    AlmanacCode/codealmanac

    A living wiki for codebases that uses AI agents to automatically extract and maintain architectural context and system knowledge in Markdown files within the repository.

  • Project
    tickernelz/opencode-mem

    A persistent memory system for AI coding agents that uses a local vector database to retain long-term project context and user preferences across sessions.

  • Project
    oceanbase/powercontext

    A context runtime for human-agent collaboration that turns shared work into persistent project context for seamless handoffs and the ability to continue work across different AI agents.