Atlas
Local code intelligence for developers and AI assistants
Start new thread

Atlas - Local code intelligence for developers and AI assistants

by•
Atlas turns a repository into a local code-intelligence graph so developers and AI coding assistants can find symbols, trace callers and dependencies, assess change impact, and retrieve focused context with exact file-and-line citations. It runs as one local binary with SQLite storage and CLI, MCP, and HTTP interfaces—no hosted Atlas service and no code leaving your machine. Published benchmarks report 36× fewer context tokens and 17× faster query retrieval versus the tested graph baseline.

Add a comment

Replies

Best
Maker
📌
We built Atlas after seeing AI coding tools consume huge prompts yet still miss the few files and relationships that actually matter. Our goal was a local map of the codebase: compact, source-grounded context for humans and assistants without sending repositories to a hosted service. Atlas indexes symbols, references, callers, routes, and likely change impact into SQLite, then returns bounded answers with file:line citations through CLI, MCP, or HTTP. We’d love feedback on language coverage, workflows, and the queries you want Atlas to answer next.