socraticode

What it does

A Docker-based MCP server that builds an automatically-maintained index of your entire codebase for semantic search and code understanding. Uses AST-aware chunking and hybrid semantic + BM25 search to enable queries, dependency graphs, impact analysis, and visual navigation. The index stays local in Docker; no data leaves your machine.

Who it's for

Backend engineers and full-stack developers working with large, multi-language codebases who want AI assistants (Claude, Cursor, VS Code) to understand code relationships, call flows, and blast radius. Multiple agents can share a single coordinated index simultaneously.

Common use cases

• Query dependency graphs and call flows to understand architectural impact of changes
• Find all callers or usages of a function or symbol across your entire codebase
• Generate searchable context artifacts (database schemas, API specs, infrastructure configs) for inclusion in prompts
• Visually navigate large repositories through the HTML graph explorer
• Index codebases up to 40 million lines of code with automatic checkpoint recovery on restart

Setup pitfalls

• Requires Docker daemon running; the MCP server cannot function without it
• First indexing pass is slow on large codebases for accuracy; subsequent changes only delta-update. Indexing is resumable across restarts
• AGPL-3.0 licensed; verify compatibility with your project's license if you are distributing code that depends on it
• Indexing requires write permissions to your project directory; the Docker container mounts volumes and writes the index to disk