Skip to main content

Travsr Documentation

The code graph that lives next to git.

Travsr builds a local graph of your code and exposes it through the Model Context Protocol (MCP). Find callers, follow dependencies, and retrieve focused context for your editor or AI assistant.

Start with your repository

Follow the Quickstart to install Travsr, index a Git repository, and connect your coding tools. The running daemon watches saved files; Git hooks reconcile commits, merges, and checkouts.

GuideWhat you'll do
QuickstartInstall, index, connect a client, and ask a question
InstallationChoose an install method and configure language analyzers
ArchitectureUnderstand indexing, graph storage, and retrieval

Explore the product

  • Source and supporting files: index programming languages alongside JSON, YAML, TOML, XML, and Markdown. Semantic coverage depends on the available language analyzer.
  • Multiple repositories: connect a global MCP server, discover registered repos with repos_list, and scope questions using repo.
  • Focused answers: retrieve symbols and source snippets within a token budget.
  • Optional local search: use travsr embed init to add embeddings for natural-language discovery. They rank candidates; they do not determine code relationships.
  • Editor workflows: inspect callers, check blast radius before editing, and copy graph context into chat from VS Code.

Reference

SectionContents
CLI ReferenceCommands, flags, and examples
MCP ToolsTool schemas, repository scope, and query examples
VS Code ExtensionFeatures, settings, and troubleshooting

How it works

  1. Index: structural parsing discovers symbols and imports. Language analyzers add semantic relationships; Markdown uses a dedicated parser.
  2. Store: a local SQLite graph records repository-scoped symbol identities and their relationships.
  3. Retrieve: graph traversal and ranking gather context, then token-budget selection keeps responses focused.
  4. Connect: use local stdio with an MCP client, or self-host the SSE server.

Analysis can continue in the background. Check indexing status and graph health when you need to verify coverage or freshness. See why follow the code graph for examples and a comparison with text search.