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. Fully local: no API key, no Docker, no telemetry.

This Python wrapper downloads the selected memory-for-ai runtime set from GitHub Releases on first run and verifies it before publishing it in your OS cache directory. The set contains the native executable and authenticated integration asset, with the graph UI always embedded.

Installation

pip install memory-for-ai
# or
pipx install memory-for-ai

There is one composition per platform: the graph UI ships in every build, so no variant selection is needed.

Usage

memory-for-ai install             # configure every detected coding agent (--dry-run previews)
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 --help

Restart your agent, say "Index this project" — done. Update with pip install -U memory-for-ai.

Supported platforms

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

Documentation

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 — measure token and tool-call savings on your own repo