Semgrep MCP Server

A Model Context Protocol (MCP) server for using Semgrep to scan code for security vulnerabilities. Secure your vibe coding! 😅

Model Context Protocol (MCP) is a standardized API for LLMs, Agents, and IDEs like Cursor, VS Code, Windsurf, or anything that supports MCP, to get specialized help, get context, and harness the power of tools. Semgrep is a fast, deterministic static analysis tool that semantically understands many languages and comes with over 5,000 rules. 🛠️

[!NOTE]
This beta project is under active development. We would love your feedback, bug reports, feature requests, and code. Join the #mcp community Slack channel!

Contents

• Semgrep MCP Server
  • Contents
  • Getting started
    • Cursor
    • ChatGPT
    • Hosted Server
      • Cursor
  • Demo
  • API
    • Tools
      • Scan Code
      • Understand Code
      • Cloud Platform (login and Semgrep token required)
      • Meta
    • Prompts
    • Resources
  • Usage
    • Standard Input/Output (stdio)
      • Python
      • Docker
    • Streamable HTTP
      • Python
      • Docker
    • Server-sent events (SSE) (deprecated)
  • Semgrep AppSec Platform
  • Integrations
    • Cursor IDE
    • VS Code / Copilot
      • Manual Configuration
      • Using Docker
    • Windsurf
    • Claude Desktop
    • Claude Code
    • OpenAI
      • Agents SDK
    • Custom clients
      • Example Python streamable HTTP client
  • Contributing, community, and running from source
    • Similar tools 🔍
    • Community projects 🌟
    • MCP server registries

Getting started

Install the Semgrep binary as described elsewhere in this repository, and use it to run the MCP server:

semgrep mcp # see --help for more options

Or, run as a Docker container:

docker run -i --rm semgrep/semgrep semgrep mcp

Cursor

Example mcp.json

{
  "mcpServers": {
    "semgrep": {
      "command": "semgrep",
      "args": ["mcp"],
      "env": {
        "SEMGREP_APP_TOKEN": "<token>"
      }
    }
  }
}

Add an instruction to your .cursor/rules to use automatically:

Always scan code generated using Semgrep for security vulnerabilities

ChatGPT

1. Go to the Connector Settings page (direct link)
2. Name the connection Semgrep
3. Set MCP Server URL to https://mcp.semgrep.ai/mcp
4. Set Authentication to No authentication
5. Check the I trust this application checkbox
6. Click Create

See more details at the official docs.

Hosted Server

[!WARNING]
mcp.semgrep.ai is an experimental server that may break unexpectedly. It will rapidly gain new functionality.🚀

Cursor

1. Cmd + Shift + J to open Cursor Settings
2. Select MCP Tools
3. Click New MCP Server.

{
  "mcpServers": {
    "semgrep": {
      "type": "streamable-http",
      "url": "https://mcp.semgrep.ai/mcp"
    }
  }
}

Demo

API

Tools

Enable LLMs to perform actions, make deterministic computations, and interact with external services.

Scan Code

• security_check: Scan code for security vulnerabilities
• semgrep_scan: Scan code files for security vulnerabilities with a given config string
• semgrep_scan_with_custom_rule: Scan code files using a custom Semgrep rule

Understand Code

• get_abstract_syntax_tree: Output the Abstract Syntax Tree (AST) of code

Cloud Platform (login and Semgrep token required)

• semgrep_findings: Fetch Semgrep findings from the Semgrep AppSec Platform API

Meta

• supported_languages: Return the list of languages Semgrep supports
• semgrep_rule_schema: Fetches the latest semgrep rule JSON Schema

Prompts

Reusable prompts to standardize common LLM interactions.

• write_custom_semgrep_rule: Return a prompt to help write a Semgrep rule

Resources

Expose data and content to LLMs

• semgrep://rule/schema: Specification of the Semgrep rule YAML syntax using JSON schema
• semgrep://rule/{rule_id}/yaml: Full Semgrep rule in YAML format from the Semgrep registry

Usage

In order to use the Semgrep MCP server, you must first have the Semgrep CLI:

$ brew install semgrep

The server can then be invoked via the mcp subcommand:

$ semgrep mcp --help

Usage: semgrep mcp [OPTIONS]

  Entry point for the MCP server

  Supports stdio and streamable-http transports. For stdio, it will read
  from stdin and write to stdout. For streamable-http, it will start
  an HTTP server on port 8000.

Options:
  -v, --version                   Show version and exit.
  -t, --transport [stdio|streamable-http]
                                  Transport protocol to use:
                                  stdio or streamable-http
  -p, --port INTEGER              Port to use for the MCP server
  -h, --help                      Show this message and exit.

Standard Input/Output (stdio)

The stdio transport enables communication through standard input and output streams. This is particularly useful for local integrations and command-line tools. See the spec for more details.

Python

semgrep mcp

By default, the server will run in stdio mode. Because it's using the standard input and output streams, it will look like the tool is hanging without any output, but this is expected.

Docker

The Semgrep binary is published to Docker:

docker run -i --rm semgrep/semgrep semgrep mcp -t stdio

Streamable HTTP

Streamable HTTP enables streaming responses over JSON RPC via HTTP POST requests. See the spec for more details.

By default, the server listens on 127.0.0.1:8000/mcp for client connections. To change any of this, set FASTMCP_* environment variables. The server must be running for clients to connect to it.

Python

semgrep mcp -t streamable-http

By default, the server will run in stdio mode, so you will have to include -t streamable-http.

Docker

docker run -p 8000:8000 semgrep/semgrep semgrep mcp

Server-Sent Events (SSE) (deprecated)

[!WARNING]
The MCP community considers this a legacy transport protocol. We have stopped supporting the SSE transport. Please use Streamable HTTP instead.

Semgrep AppSec Platform

Optionally, to connect to Semgrep AppSec Platform:

1. Login or sign up
2. Generate a token from Settings
3. Add the token to your environment variables:

   • CLI (export SEMGREP_APP_TOKEN=<token>)

   • Docker (docker run -e SEMGREP_APP_TOKEN=<token>)

   • MCP config JSON

    "env": {
      "SEMGREP_APP_TOKEN": "<token>"
    }

[!TIP]
Please reach out for support if needed. ☎️

Integrations

Cursor IDE

1. Install Semgrep:

   brew install semgrep
   # or
   python3 -m pip install semgrep

2. Authenticate and install Semgrep Pro:

   semgrep login && semgrep install-semgrep-pro

3. Add the following JSON block to your ~/.cursor/mcp.json global or .cursor/mcp.json project-specific configuration file:

   {
     "mcpServers": {
       "semgrep": {
         "command": "semgrep mcp",
         "env": {},
         "args": []
       }
     }
   }

4. Create a .cursor/hooks.json file in your project to enable automatic scanning:

   {
     "version": 1,
     "hooks": {
       "stop": [{"command": "semgrep mcp -k stop-cli-scan -a cursor"}],
       "afterFileEdit": [{"command": "semgrep mcp -k record-file-edit -a cursor"}]
     }
   }

See cursor docs for more info.

VS Code / Copilot

Click the install buttons at the top of this README for the quickest installation.

Manual Configuration

Add the following JSON block to your User Settings (JSON) file in VS Code. You can do this by pressing Ctrl + Shift + P and typing Preferences: Open User Settings (JSON).

{
  "mcp": {
    "servers": {
      "semgrep": {
        "command": "semgrep",
        "args": ["mcp"]
      }
    }
  }
}

Optionally, you can add it to a file called .vscode/mcp.json in your workspace:

{
  "servers": {
    "semgrep": {
      "command": "semgrep",
        "args": ["mcp"]
    }
  }
}

Using Docker

{
  "mcp": {
    "servers": {
      "semgrep": {
        "command": "docker",
        "args": [
          "run",
          "-i",
          "--rm",
          "semgrep/semgrep",
          "semgrep",
          "mcp",
          "-t",
          "stdio"
        ]
      }
    }
  }
}

See VS Code docs for more info.

Windsurf

Add the following JSON block to your ~/.codeium/windsurf/mcp_config.json file:

{
  "mcpServers": {
    "semgrep": {
      "command": "semgrep",
      "args": ["mcp"]
    }
  }
}

See Windsurf docs for more info.

Claude Desktop

Here is a short video showing Claude Desktop using this server to write a custom rule.

Add the following JSON block to your claude_desktop_config.json file:

{
  "mcpServers": {
    "semgrep": {
      "command": "semgrep",
      "args": ["mcp"]
    }
  }
}

See Anthropic docs for more info.

Claude Code

1. Install Semgrep:

   brew install semgrep
   # or
   python3 -m pip install semgrep

2. Launch Claude Code in your terminal:

   claude

3. Add the marketplace source:

   /plugin marketplace add semgrep/mcp-marketplace

4. Install the plugin:

   /plugin install semgrep-plugin@semgrep

5. Configure the plugin:

   /semgrep-plugin:setup_semgrep_plugin

   (If that fails, try /plugin enable semgrep-plugin@semgrep)

See Claude Code docs for more info.

OpenAI

See the official docs:

• https://platform.openai.com/docs/mcp
• https://platform.openai.com/docs/guides/tools-remote-mcp

Agents SDK

async with MCPServerStdio(
    params={
        "command": "semgrep",
        "args": ["mcp"],
    }
) as server:
    tools = await server.list_tools()

See OpenAI Agents SDK docs for more info.

Custom clients

Example Python streamable HTTP client

import asyncio
import json
from mcp.client.session import ClientSession
from mcp.client.streamable_http import streamablehttp_client


async def main():
    async with streamablehttp_client("http://localhost:8000/mcp") as (read_stream, write_stream, _):
        async with ClientSession(read_stream, write_stream) as session:
            await session.initialize()
            results = await session.call_tool(
                "semgrep_scan_remote",
                {
                    "code_files": [
                        {
                            "path": "hello_world.py",
                            "content": "def hello(): print('Hello, World!')",
                        }
                    ]
                },
            )
            content_block = results.content[0]
            content = json.loads(content_block.text)
            paths = content.get("paths", None)
            if paths:
                scanned = paths.get("scanned", [])
                findings = content.get("results", [])
                print(f"Scanned {len(scanned)} paths. Found {len(findings)} findings.")

[!TIP]
Some client libraries want the URL: http://localhost:8000/mcp
and others only want the HOST: localhost:8000.
Try out the URL in a web browser to confirm the server is running, and there are no network issues.
Set SEMGREP_IS_HOSTED=true to use the semgrep_scan_remote tool

See official SDK docs for more info.

Contributing, community, and running from source

[!NOTE]
We love your feedback, bug reports, feature requests, and code. Join the #mcp community Slack channel!

See CONTRIBUTING.md for more info and details on how to run from the MCP server from source code.

Similar tools 🔍

• semgrep-vscode - Official VS Code extension
• semgrep-intellij - IntelliJ plugin

Community projects 🌟

• semgrep-rules - The official collection of Semgrep rules
• mcp-server-semgrep - Original inspiration written by Szowesgad and stefanskiasan

MCP server registries

• Glama

• MCP.so

Made with ❤️ by the Semgrep Team