memory-for-ai

An MCP server that turns a codebase into a persistent knowledge graph — functions, classes, call chains, HTTP routes, cross-service links — so an AI coding agent answers structural questions with graph queries instead of reading file after file. This npm wrapper downloads, verifies, and caches the native runtime set for your platform; Node.js owns download, cache repair, and launch.

One self-contained native executable behind the wrapper. 162 languages via vendored tree-sitter grammars, refined by embedded Hybrid-LSP type resolution. 22 MCP tools. No Docker, no API key, no telemetry — everything runs locally.

• AI agents: read docs/AGENT_GUIDE.md — the complete operating manual.
• Evaluating effectiveness: docs/MEASURING.md — a 15-minute spot check and a full A/B protocol.

Installation

npm install -g memory-for-ai-mcp

The runtime set for your platform is downloaded and verified automatically at install time. One composition per platform; the graph UI is always included (the former CBM_VARIANT=ui opt-in is obsolete).

Then configure your coding agents and index:

memory-for-ai install      # configures all 45 supported automatic/conditional client surfaces (--dry-run previews)

Claude Code, Codex CLI, Gemini CLI, Zed, OpenCode, Antigravity, Aider, KiloCode, VS Code, Cursor, Windsurf, Augment / Auggie, OpenClaw, Kiro, Junie, Hermes, OpenHands, Cline, Warp, Qwen Code, GitHub Copilot CLI, Factory Droid, Crush, Goose, Mistral Vibe, Grok Build, Qoder CLI, Kimi Code CLI, GitLab Duo CLI, Rovo Dev CLI, Amp, Devin CLI / Local, Tabnine, Continue / cn (conditional), Visual Studio (conditional, Windows), TRAE (conditional), Roo Code (conditional), Amazon Q Developer IDE, CodeBuddy Code CLI, IBM Bob IDE (conditional), IBM Bob Shell, Pochi, Pi, Sourcegraph Cody (explicit opt-in), Oh My Pi (omp).

Restart your agent. Say "Index this project" — done.

Usage

memory-for-ai install          # configure all detected coding agents
memory-for-ai install --project  # repo-local fenced server named after the repo (no global changes)
memory-for-ai cli list_projects   # every MCP tool also runs as a one-shot CLI command
memory-for-ai --version
memory-for-ai --help
memory-for-ai uninstall        # remove owned agent configs and the binary

Updates go through npm on every platform: npm install -g memory-for-ai-mcp-mcp@latest. The command stays memory-for-ai (the package name differs because memory-for-ai on npm is held by an unrelated third party).

Supported Platforms

| OS | Architecture | |---------|-------------| | macOS | arm64, amd64 | | Linux | arm64, amd64 | | Windows | arm64, amd64 |

Learn more

• Agent operating guide — tool catalog, task→tool playbooks, correctness protocol, per-project tuning
• Installation reference — every install path, CI/containers, build from source
• Configuration reference — settings and environment variables
• Measuring real effectiveness — tokens, tool calls, answer quality on your repo

License

MIT