Atlas documentation

Local code intelligence for developers and AI coding assistants. Install Atlas, index a repository, and return focused, source-grounded context through the CLI, MCP, and a local HTTP service.

• Getting Started — Create a local Atlas index, run the first cited queries, and connect an AI coding assistant.
• Installation — Install Atlas via Homebrew, npm, release archives, or native Linux packages, then verify it.
• Core Concepts — The Atlas mental model: the code knowledge graph, snapshots, workspace resolution, output control, and the CLI, HTTP, and MCP surfaces.
• Indexing and Reindexing — Keep the Atlas index fresh: incremental updates, watch mode, exclusions, and recovery.
• CLI Reference — Atlas command-line reference: search, symbols, relationships, impact, routes, and maintenance.
• AI Assistant Setup — Connect Claude, Codex, and other MCP-compatible assistants to Atlas.
• MCP Tools — Atlas MCP tools that give coding assistants bounded, cited code context.
• Dashboard and HTTP API — Run the local Atlas dashboard and the HTTP / MCP service.
• Configuration — Configure Atlas limits, storage location, and runtime behavior.
• Privacy and Data Handling — How Atlas stores repository intelligence locally and what stays on your machine.
• Supported Languages and Formats — Languages and formats Atlas indexes, with capability levels.
• Benchmarks and Methodology — How to read and reproduce the published Atlas benchmark evidence.
• Troubleshooting — Diagnose stale, missing, or incomplete Atlas results.
• Upgrade and Uninstall — Update Atlas safely, or remove it and its local data.