SocratiCode

No global install required — just add the npx entry to your MCP host's config file. SocratiCode downloads on first use and manages its own Docker-based backend services.

{
  "mcpServers": {
    "socraticode": {
      "command": "npx",
      "args": ["-y", "socraticode"]
    }
  }
}

By default SocratiCode uses Ollama running in Docker for fully local embeddings. To use OpenAI or Google embeddings instead, set environment variables in your MCP config.

{
  "mcpServers": {
    "socraticode": {
      "command": "npx",
      "args": ["-y", "socraticode"],
      "env": {
        "EMBEDDING_PROVIDER": "openai",
        "EMBEDDING_MODEL": "text-embedding-3-small",
        "OPENAI_API_KEY": "sk-your-key-here"
      }
    }
  }
}

Open your AI assistant and ask it to index the current project. SocratiCode will crawl the workspace, generate embeddings, and build the dependency graph. Large codebases may take several minutes; indexing is resumable if interrupted.

Check the indexing status by asking your AI assistant to report progress. The codebase_status tool returns the current number of indexed files, embedding progress, and any errors.

After indexing completes, ask the assistant to build the code graph. This performs polyglot static analysis to construct the import/call/dependency graph for impact analysis.

Set SOCRATICODE_BRANCH_AWARE=true to maintain separate vector collections per git branch. Each branch gets its own index, so searches always reflect the current branch's code.

{
  "mcpServers": {
    "socraticode": {
      "command": "npx",
      "args": ["-y", "socraticode"],
      "env": {
        "SOCRATICODE_BRANCH_AWARE": "true",
        "EMBEDDING_PROVIDER": "ollama"
      }
    }
  }
}