SocratiCode is a fairly comprehensive local codebase-intelligence MCP server offering hybrid semantic search, polyglot dependency graphs, and symbol-level impact analysis, running by default in local Docker containers with no API key required — well suited for deep exploration of large, multi-language repositories. Adopting it means accepting locally-run Docker containers, optional cloud API keys for alternate embedding providers, and AGPL-3.0 licensing terms; the documentation is explicit that its static-analysis call graph has blind spots around dynamic dispatch and framework magic, so impact-analysis results should be paired with manual review.
Strongest · Documentation 18/20 Weakest · Reliability 9/20
Static review · not runListed 2026-08-06
Read the FMRS scoring method →Runtime：Node.js 18+ · Docker
OPENAI_API_KEY
Needed only when EMBEDDING_PROVIDER=openai; get one at platform.openai.com.
GOOGLE_API_KEY
Needed only when EMBEDDING_PROVIDER=google; get one at aistudio.google.com/apikey.
QDRANT_API_KEY
API key for the remote Qdrant instance, needed only when QDRANT_MODE=external.
LMSTUDIO_API_KEY
API key if auth is enabled in LM Studio.
LITELLM_API_KEY
LiteLLM proxy master/virtual key, required in LiteLLM mode.
EMBEDDING_PROVIDER
Embedding provider: ollama (default), openai, google, lmstudio, or litellm; optional.
OLLAMA_URL
URL of an external Ollama server (default auto-detected, e.g. http://localhost:11434).
OLLAMA_MODE
Set to external to use a native or remote Ollama; default auto.
EMBEDDING_MODEL
Embedding model name; defaults per provider, required for LM Studio/LiteLLM modes.
EMBEDDING_DIMENSIONS
Embedding vector dimensions; required for LM Studio/LiteLLM modes.
QDRANT_MODE
managed (default Docker-managed) or external (user-provided instance).
QDRANT_URL
Full URL of remote/cloud Qdrant, needed only when QDRANT_MODE=external.
LMSTUDIO_URL
LM Studio local server URL, default http://localhost:1234/v1.
LITELLM_URL
LiteLLM proxy URL, default http://localhost:4000/v1.
LITELLM_SEND_DIMENSIONS
Whether to forward the dimensions parameter to the proxy; only safe for Matryoshka-aware backends.
SOCRATICODE_LOG_LEVEL
Log level for structured logging.
SOCRATICODE_PROJECT_ID
Project ID to share one index across directories/git worktrees.
SOCRATICODE_LINKED_PROJECTS
Comma-separated linked project paths for cross-project search.
SOCRATICODE_BRANCH_AWARE
Set true to maintain separate indexes per git branch.
RESPECT_GITIGNORE
Set false to disable .gitignore processing.
INCLUDE_DOT_FILES
Set true to include dot-directories (e.g. .agent).
EXTRA_EXTENSIONS
Include non-standard file extensions (e.g. .tpl), indexed as plaintext.
EXTENSION_LANGUAGE_MAP
Map a custom extension to a real language (e.g. .inc:php) for AST processing.
SEARCH_MIN_SCORE
Minimum search score threshold (default 0.10); may need lowering for cross-project search.
1) Make sure Docker is installed and running, and Node.js 18+ is available. 2) Simplest route: add the npx launch command to your MCP client config — {"command":"npx","args":["-y","socraticode"]} under mcpServers (Claude Desktop, Windsurf, Cline, Roo Code) or servers (VS Code project-local .vscode/mcp.json). 3) Claude Code users are recommended to install the official plugin, which bundles workflow skills: run claude plugin marketplace add giancarloerra/socraticode then claude plugin install socraticode@socraticode (or the equivalent /plugin marketplace add and /plugin install slash commands inside Claude Code). 4) VS Code/Cursor users can also install the dedicated extension/plugin from the VS Code Marketplace, Open VSX, or Cursor Marketplace. 5) On first use the server automatically pulls Docker images and starts Qdrant and Ollama containers and downloads the embedding model — a one-time step taking roughly 5 minutes; subsequent starts take seconds. 6) For cloud embeddings (OpenAI/Google) or an external Qdrant instance, set env vars such as EMBEDDING_PROVIDER, OPENAI_API_KEY, GOOGLE_API_KEY, QDRANT_MODE, QDRANT_URL, and QDRANT_API_KEY in the client config. 7) On a project for the first time, ask the AI to 'index this codebase' to trigger codebase_index, then check progress with codebase_status.
{"mcpServers":{"socraticode":{"command":"npx","args":["-y","socraticode"]}}}
Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.
{
"servers": {
"socraticode": {
"command": "npx",
"args": [
"-y",
"socraticode"
]
}
}
}
Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).
claude mcp add socraticode -- npx -y socraticode
Run it in a terminal; replace any <…> placeholders with your own values first.
After starting the host, the client tool list should show codebase_index, codebase_search, and the other SocratiCode tools; say "Index this codebase" on first use, then poll codebase_status until it reports the index is complete — that proves the connection works.
Once connected, you can ask your AI assistant things like:
No matching tools
Listed from the project's documentation, not tested by this site.
SocratiCode is an open-source MCP server that gives AI assistants deep semantic understanding of a codebase. It runs hybrid semantic + BM25 (RRF-fused) search on top of Qdrant (vector database) and Ollama (local embeddings by default, switchable to OpenAI or Google), uses ast-grep for AST-aware chunking across 18+ languages, and builds a polyglot dependency graph with symbol-level impact analysis (blast radius), call-flow tracing, circular-dependency detection, and an interactive HTML graph viewer. It also supports cross-project and branch-aware search, and indexing of non-code knowledge such as database schemas, API specs, and infra configs. The README reports, from a benchmark on VS Code's 2.45M-line codebase, 61% less context, 84% fewer tool calls, and 37x faster responses versus grep-based exploration. By default everything runs locally via Docker with no API keys required and no code leaving the machine; cloud embedding providers and external Qdrant instances are optional. The project is licensed under AGPL-3.0 and also offers a separate paid cloud offering (SocratiCode Cloud) currently in private beta.
Upstash's official server providing up-to-date third-party library docs for AI coding assistants
AI-powered NPM package analysis MCP server
An independent control layer that gives AI coding agents budgets, brakes, and receipts
GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language
Source revision 4aa6fc23ae69 Data synced 2026-11-04 Read the FMRS scoring method