
SocratiCode MCP Server: Setup for Claude, Cursor and Other Apps
Lets Claude, Cursor and other assistants search your code by meaning and see how files, functions and imports connect.
This is a community server by Giancarlo Erra. It runs on your computer and builds a searchable index of your project. By default it needs Docker, which runs a local vector database and an embedding model, and no API key.
You can switch the embeddings to OpenAI or Google, or use a hosted Qdrant database. If you do, parts of your code leave your machine.
Good for
- Developers who work in large or unfamiliar codebases
- Engineers who want to know what a change could break
- Teams who want code search that stays on their own machine
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install SocratiCode MCP
Claude Code
- 1
Copy this prompt.
Add the SocratiCode MCP server to Claude Code. Run this command: claude mcp add socraticode -- npx -y socraticode Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js, Docker running (default local Qdrant and Ollama).
Docker must be running before the first use: the default setup pulls images and starts a local Qdrant (vector store) and Ollama (embeddings) container, and the README says Docker is required in the default managed mode. Index a project by asking the assistant, then check the index status until it finishes. No API key is needed for the default setup; the optional embedding provider variables are listed in the README.
Built from the package data we hold. Not yet tested by us.
Tools
Starts indexing a codebase for search
Begins indexing a project in the background and returns at once. Check the status tool until it reaches 100% before searching.
Try: “Index this codebase.”
Runs code ·
codebase_indexShows indexing progress
Reports chunk count, progress, the last finished operation and file watcher state.
Try: “What is the codebase index status?”
Read ·
codebase_statusSearches the code by meaning
Runs a semantic search over an indexed codebase from a plain-language question and returns matching code chunks.
Try: “Where do we validate login tokens?”
Read ·
codebase_searchRefreshes the index with changed files
Re-indexes only the files that changed, and waits until it is done.
Try: “Update the index with my latest edits.”
Write ·
codebase_updateKeeps the index up to date as files change
Starts or stops watching a folder, and catches up with an incremental update first when it starts.
Try: “Watch this project and keep the index current.”
Write ·
codebase_watchStops indexing that is running
Stops an index run after the current batch, keeping progress so it can resume.
Try: “Stop the indexing for now.”
Write ·
codebase_stopShows what could break if you change something
Lists the files and functions that depend on a file or symbol, so you can see the blast radius of a change.
Try: “What breaks if I change validateUser?”
Read ·
codebase_impactTraces what code calls from an entry point
Without arguments, lists likely entry points. With one, returns the call tree from it.
Try: “Trace the flow starting at main.”
Read ·
codebase_flowShows a function or class in full
Gives the definition, kind, callers and callees of one symbol, with confidence levels.
Try: “Show me everything about parseConfig.”
Read ·
codebase_symbolLists symbols in a file or finds them by name
Lists the symbols in a file, or searches the project for a name.
Try: “List the functions in src/auth.ts.”
Read ·
codebase_symbolsBuilds a map of how files import each other
Builds a dependency graph by static analysis of import and export links, in the background.
Try: “Build the dependency graph.”
Write ·
codebase_graph_buildShows dependency graph build progress
Reports build progress, node and edge counts, and when the graph was last built.
Try: “Is the dependency graph ready?”
Read ·
codebase_graph_statusShows what a file imports and what depends on it
Returns the imports of a file and the files that depend on it.
Try: “What depends on src/db.ts?”
Read ·
codebase_graph_querySummarizes the dependency graph
Returns file and edge counts, the most connected files, orphan files and circular dependencies.
Try: “Give me stats on the dependency graph.”
Read ·
codebase_graph_statsFinds circular dependencies
Lists cycles of files that import each other.
Try: “Are there circular dependencies?”
Read ·
codebase_graph_circularDraws the dependency graph
Returns a Mermaid diagram as text, or writes an interactive HTML page and opens it in your browser.
Try: “Show the dependency graph as a diagram.”
Files ·
codebase_graph_visualizeDeletes the saved dependency graph
Removes a project's stored graph. It can be rebuilt later.
Try: “Remove the saved graph for this project.”
Delete ·
codebase_graph_removeLists extra project knowledge you defined
Lists the artifacts in the context file, such as database schemas and API specs, with their index status.
Try: “What context artifacts does this project have?”
Read ·
codebase_contextSearches schemas, API specs and docs by meaning
Runs a semantic search over the context artifacts, indexing them on first use.
Try: “Which tables hold customer addresses?”
Write ·
codebase_context_searchIndexes the context artifacts
Chunks and embeds the artifacts so they can be searched. Usually not needed because search does it.
Try: “Index the context artifacts.”
Write ·
codebase_context_indexDeletes the indexed context artifacts
Removes a project's indexed context artifacts from the vector database.
Try: “Remove the context index.”
Delete ·
codebase_context_removeDeletes a project's code index
Removes the whole index for a project from the vector database, after stopping the watcher and any running index work. Your source files are not touched.
Try: “Remove the index for this project.”
Delete ·
codebase_removeReports or cleans up stored project data
Lists stored project identities and their resources. It only reports unless you confirm an exact identity with a fresh token.
Try: “Show me stored projects that may be abandoned.”
Delete ·
codebase_pruneLists indexed projects
Lists every project that has an index in the vector database.
Try: “Which projects are indexed?”
Read ·
codebase_list_projectsChecks Docker, Qdrant and the embedding model
Diagnoses the supporting services the server needs.
Try: “Why is code search not working?”
Read ·
codebase_healthExplains what SocratiCode is
Returns an overview of the server and its tools.
Try: “What can SocratiCode do?”
Read ·
codebase_about
Read from the source code
Permissions
medium risk · 4 things to watch
Medium risk
It reads your project files and keeps an index of them, and some tools delete that index. The default setup stays on your machine, so check where your code goes if you switch to cloud embeddings.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It needs Docker and starts containers
First use pulls images and starts Qdrant and Ollama through Docker. Make sure you are happy to run them. github.com
Cloud embeddings send code out
Choosing OpenAI, Google or a remote Qdrant sends code chunks to those services. Keep the default local setup for private code. github.com
Some tools delete saved indexes
Remove and prune tools delete stored index data, not your source files. Check the target before approving. github.com
The graph viewer can open your browser
The interactive mode writes an HTML page and opens it. Ask for the text diagram if you do not want that. github.com
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Talks to | Local Docker services by default; OpenAI, Google or a remote Qdrant only if you configure them |
|---|---|
| Sign-in and keys | No key for the default local setup. Optional OpenAI, Google, Qdrant, LM Studio or LiteLLM keys are read from environment variables. |
Questions
Is SocratiCode official?
No. It is a community project by Giancarlo Erra and is not a product of any code host or editor maker.
Do I need an API key?
No. The default local setup needs none. OpenAI, Google, LM Studio, LiteLLM and hosted Qdrant keys are optional.
What do I need installed?
Node.js 18.17 or newer and Docker running, which the README says the default local stack needs.
Does my code leave my machine?
Not with the default setup. The README says it is private and local by default, and sending code out only happens if you configure cloud embeddings or a remote Qdrant.
Is it free?
The package is AGPL-3.0, and the repository also ships a commercial license file. Check the license terms for your use.
How do I index my project?
Ask the assistant to index the codebase, then ask for the index status until it is complete. Search only works after that.
Information
updated Nov 2, 2026
- Publisher
- Community
- Package
socraticode- Registry
- npm
- Transport
- stdio
- Auth
- None
- Version
- 1.16.1
- License
- AGPL-3.0
- Last commit
- Oct 23, 2026
- Open issues
- 1
- Listed in
- Official MCP registry
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.
Show all 2Hide
- README.md · github.com
- socraticode · npmjs.com