Codebase

See what your repository is actually made of

Indexing parses the repository with tree-sitter, extracts symbols and the relationships between them, then ranks the graph by PageRank so the code everything else depends on rises to the top. The canvas opens on components, drills into modules, and ends on a symbol graph you can read. Indexing is on by default and can be disabled per repository.

• TypeScript, TSX and Python, parsed, not pattern-matched
• Files, classes, functions, methods and interfaces as first-class symbols
• contains, imports, calls, extends and implements edges
• Call edges carry a confidence tier, and the canvas can hide the guesses
• PageRank orders the map, so the load-bearing code is visible first
• Re-indexed on push to the default branch, on a nightly sweep, or on demand
• The graph is ready before embeddings finish, so the map never waits on search

3languages parsed

1. Semantic code searchAgents find code by describing what it does. Retrieval is by embedding similarity, ranked by the structure of the graph.