Planning-specialized knowledge graph for Markdown-driven projects
GitContext parses Markdown planning documents (PLAN.md, TASK.md, README.md, knowledge/*) into a structured knowledge graph — decisions, tasks, phases, patterns, and their dependencies. Exposed as an MCP server for AI editors like Claude Code, Cursor, and Windsurf.
"What does this decision affect?" "What blocks this task?" "Show me all orphaned decisions." — answered from the graph, not grep.
Built for AI editors. Your Claude Code / Cursor session queries the planning graph directly. Zero-token context switching.
Re-indexes on every commit. Only changed files. Sub-500ms incremental updates. Branch-aware graph namespaces.
Single static binary. Rust. KuzuDB embedded. No Node.js, no Python, no Docker required for CLI use.
Based on the GitCortex architecture (6-crate Rust workspace, tree-sitter, KuzuDB, MCP server, WebGL viz). Currently adapting the "code graph" paradigm to "planning graph."