AutoDocs

Sita automatically generates and maintains technical documentation for your codebase by parsing repositories via AST and SCIP, building a topological dependency graph, and using an LLM to write docs for each file/definition in dependency order. It also provides a search agent (Martin) that retrieves only the relevant code context and hands it to AI coding tools, reducing token usage by 40–60%. Designed for engineering teams using agentic coding tools like Cursor, Claude Code, or Cline, Sita integrates via MCP to feed agents the exact files and functions they need. It works with any codebase, including monorepos, and supports incremental updates via Merkle tree diffing. What makes Sita different is its focus on dependency-aware context: instead of whole-codebase prompts, it sends minimal, cited context. The generated docs are served through a visual React UI with dependency graphs, and the system automatically refreshes on push to main. It is open source, self-hostable, or available as a managed service for business. Sita also aims to prevent cascading changes by flagging impacted dependents and upstream contracts, helping teams avoid surprise regressions. The roadmap includes DFS/BFS graph traversal for more accurate change impact analysis.