Stay focused on building your product while your AI assistant handles the security heavy lifting with GitGuardian's comprehensive protection.
This MCP server enables your AI agent to scan projects using GitGuardian's industry-leading API, featuring over 500 secret detectors to prevent credential leaks before they reach public repositories.
Resolve security incidents without context switching to the GitGuardian console. Take advantage of rich contextual data to enhance your agent's remediation capabilities, enabling rapid resolution and automated removal of hardcoded secrets.
Caution
MCP servers are an emerging and rapidly evolving technology. While they can significantly boost productivity and improve the developer experience, their use with various agents and models should always be supervised.
Agents act on your behalf and under your responsibility. Always use MCP servers from trusted sources (just as you would with any dependency), and carefully review agent actions when they interact with MCP server tools.
To better assist you in safely using this server, we have:
(1) Designed our MCP server to operate with "read-only" permissions, minimizing the access level granted to your agent. This helps ensure that, even if the agent tries to perform unintended actions, its capabilities remain limited to safe, non-destructive operations.
(2) Released this official MCP server to ensure you are using a legitimate and trusted implementation.
- Secret Scanning: Scan code for leaked secrets, credentials, and API keys
- Incident Management: View security incidents related to the project you are currently working.
- Honeytokens: Create honeytokens to detect unauthorized access
- Authentication Management: Get authenticated user information and token details
- Token Management: Revoke current API tokens
Want more features? Have a use case that's not covered? We'd love to hear from you! Submit your ideas and feedback by opening an issue on GitHub to help us prioritize new MCP server capabilities.
Remediate all incidents related to my project
Scan this codebase for any leaked secrets or credentials
Check if there are any new security incidents assigned to me
Help me understand this security incident and provide remediation steps
List all my active honeytokens
Generate a new honeytoken for monitoring AWS credential access
Show me my most recent honeytoken and help me embed it in my codebase
Create a honeytoken named 'dev-database' and hide it in config files
Before installing the GitGuardian MCP servers, ensure you have the following prerequisites:
- uv: This project uses uv for package installation and dependency management. Install uv by following the instructions at: https://docs.astral.sh/uv/getting-started/installation/
Below are instructions for installing the GitGuardian MCP servers with various AI editors and interfaces.
The MCP server supports both GitGuardian SaaS and self-hosted instances.
Quick Install with One-Click Buttons (Cursor >= 1.0):
For Developer MCP Server:
Note: The one-click install sets up the default US SaaS configuration. For EU SaaS or self-hosted instances, you'll need to manually add environment variables as shown in the Configuration section.
Manual Configuration:
-
Edit your Cursor MCP configuration file located at
~/.cursor/mcp.json -
Add the GitGuardian MCP server configuration:
{ "mcpServers": { "GitGuardianDeveloper": { "command": "uvx", "args": [ "--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server" ] } } }
-
Edit your Claude Desktop MCP configuration file located at:
- macOS:
~/Library/Application Support/Claude Desktop/mcp.json - Windows:
%APPDATA%\Claude Desktop\mcp.json
- macOS:
-
Add the GitGuardian MCP server configuration:
{ "mcpServers": { "GitGuardianDeveloper": { "command": "/path/to/uvx", "args": [ "--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server" ] } } } -
Replace
/path/to/uvxwith the absolute path to the uvx executable on your system.⚠️ WARNING: For Claude Desktop, you must specify the full absolute path to theuvxexecutable, not just"command": "uvx". This is different from other MCP clients. -
Restart Claude Desktop to apply the changes.
To use the GitGuardian MCP server with Windsurf:
-
Edit your Windsurf MCP configuration file located at:
- macOS:
~/Library/Application Support/Windsurf/mcp.json - Windows:
%APPDATA%\Windsurf\mcp.json - Linux:
~/.config/Windsurf/mcp.json
- macOS:
-
Add the following entry to the configuration file:
{ "mcp": { "servers": { "GitGuardianDeveloper": { "type": "stdio", "command": "uvx", "args": [ "--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server" ] } } } }
-
Edit your Zed MCP configuration file located at:
- macOS:
~/Library/Application Support/Zed/mcp.json - Linux:
~/.config/Zed/mcp.json
- macOS:
-
Add the GitGuardian MCP server configuration:
{ "GitGuardianDeveloper": { "command": { "path": "uvx", "args": [ "--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server" ] } } }
The GitGuardian MCP server supports multiple authentication methods depending on your deployment mode.
When using stdio transport (the default for desktop IDE integrations), the server uses OAuth for authentication by default:
- OAuth is enabled by default (
ENABLE_LOCAL_OAUTH=true) for local-first usage - When you start the server, it will automatically open a browser window to authenticate with GitGuardian
- After you log in to GitGuardian and authorize the application, you'll be redirected back to the local server
- The authentication token will be securely stored in
~/.gitguardian/for future use - The next time you start the server, it will reuse the stored token without requiring re-authentication
Example configuration (OAuth is enabled by default, no need to specify):
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/GitGuardian/ggmcp.git",
"developer-mcp-server"
]
}
}
}To disable OAuth (e.g., for using PAT instead):
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/GitGuardian/ggmcp.git",
"developer-mcp-server"
],
"env": {
"ENABLE_LOCAL_OAUTH": "false",
"GITGUARDIAN_PERSONAL_ACCESS_TOKEN": "your_pat_here"
}
}
}
}For non-interactive environments, CI/CD pipelines, or when you prefer not to use OAuth, you can authenticate using a Personal Access Token:
- Create a Personal Access Token in your GitGuardian dashboard
- Set the
GITGUARDIAN_PERSONAL_ACCESS_TOKENenvironment variable
Example configuration with PAT:
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/GitGuardian/ggmcp.git",
"developer-mcp-server"
],
"env": {
"GITGUARDIAN_PERSONAL_ACCESS_TOKEN": "your_personal_access_token_here"
}
}
}
}When using HTTP/SSE transport (with MCP_PORT set), the server expects authentication via the Authorization header in each HTTP request. This is the recommended approach for server deployments.
Important: Since ENABLE_LOCAL_OAUTH defaults to true, you must explicitly set it to false when using HTTP/SSE mode:
# Start server with HTTP transport (OAuth must be disabled)
ENABLE_LOCAL_OAUTH=false MCP_PORT=8000 MCP_HOST=127.0.0.1 uvx --from git+https://github.com/GitGuardian/ggmcp.git developer-mcp-server
# Make authenticated request
curl -X POST http://127.0.0.1:8000/tools/list \
-H "Authorization: Bearer YOUR_PERSONAL_ACCESS_TOKEN" \
-H "Content-Type: application/json" \
-d '{}'Configuration validation: The server will raise an error if both MCP_PORT and ENABLE_LOCAL_OAUTH=true are set, as HTTP/SSE mode requires per-request authentication for security reasons.
The MCP server uses OAuth authentication and defaults to GitGuardian SaaS (US region) at https://dashboard.gitguardian.com. For other instances, you'll need to specify the URL:
The following environment variables can be configured:
| Variable | Description | Default | Example |
|---|---|---|---|
GITGUARDIAN_URL |
GitGuardian instance URL | https://dashboard.gitguardian.com |
https://dashboard.eu1.gitguardian.com |
GITGUARDIAN_CLIENT_ID |
OAuth client ID | ggshield_oauth |
my-custom-oauth-client |
GITGUARDIAN_SCOPES |
OAuth scopes to request | Auto-detected based on instance type | scan,incidents:read,sources:read,honeytokens:read,honeytokens:write |
GITGUARDIAN_TOKEN_NAME |
Name for the OAuth token | Auto-generated based on server type | "Developer MCP Token" |
GITGUARDIAN_TOKEN_LIFETIME |
Token lifetime in days | 30 |
60 or never |
GITGUARDIAN_PERSONAL_ACCESS_TOKEN |
Personal Access Token for authentication (alternative to OAuth) | Not set | YOUR_PAT_TOKEN |
ENABLE_LOCAL_OAUTH |
Enable local OAuth flow (stdio mode only, cannot be used with MCP_PORT) |
true (enabled by default for local-first usage) |
false |
MCP_PORT |
Port for HTTP/SSE transport (when set, enables HTTP transport instead of stdio, requires ENABLE_LOCAL_OAUTH=false) |
Not set (uses stdio) | 8000 |
MCP_HOST |
Host address for HTTP/SSE transport | 127.0.0.1 |
0.0.0.0 |
By default, the MCP server uses stdio transport for local IDE integrations. If you need to expose the MCP server over HTTP (for remote access or custom integrations), you can use the MCP_PORT and MCP_HOST environment variables.
To enable HTTP/SSE transport, set the MCP_PORT environment variable. Important: You must also set ENABLE_LOCAL_OAUTH=false since OAuth defaults to enabled:
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/GitGuardian/ggmcp.git",
"developer-mcp-server"
],
"env": {
"ENABLE_LOCAL_OAUTH": "false",
"MCP_PORT": "8000",
"MCP_HOST": "127.0.0.1"
}
}
}
}You can also run the server directly with HTTP transport:
# Run with HTTP transport (must disable OAuth)
ENABLE_LOCAL_OAUTH=false MCP_PORT=8000 MCP_HOST=127.0.0.1 uvx --from git+https://github.com/GitGuardian/ggmcp.git developer-mcp-serverThe server will automatically start on http://127.0.0.1:8000 and be accessible for remote integrations.
When using HTTP/SSE transport, authentication is done via the Authorization header on each request. See the Per-Request Authentication section for detailed configuration.
Supported header formats:
Authorization: Bearer <token>Authorization: Token <token>Authorization: <token>
Example using curl:
# List available tools
curl -X POST http://127.0.0.1:8000/tools/list \
-H "Authorization: Bearer YOUR_PERSONAL_ACCESS_TOKEN" \
-H "Content-Type: application/json" \
-d '{}'
# Call a tool
curl -X POST http://127.0.0.1:8000/tools/call \
-H "Authorization: Bearer YOUR_PERSONAL_ACCESS_TOKEN" \
-H "Content-Type: application/json" \
-d '{"name": "get_authenticated_user_info", "arguments": {}}'Example using Python:
import httpx
headers = {
"Authorization": "Bearer YOUR_PERSONAL_ACCESS_TOKEN",
"Content-Type": "application/json"
}
async with httpx.AsyncClient() as client:
response = await client.post(
"http://127.0.0.1:8000/tools/list",
headers=headers,
json={}
)
tools = response.json()Authentication Priority:
When using HTTP transport, the authentication priority is:
- Authorization header (if present in the HTTP request) - recommended for HTTP/SSE mode
- GITGUARDIAN_PERSONAL_ACCESS_TOKEN environment variable - fallback option
Note that OAuth (ENABLE_LOCAL_OAUTH=true) is not supported in HTTP/SSE mode for security reasons. Each HTTP request must include its own authentication credentials.
Notes:
uvicornis included as a dependency - no additional installation needed.- When
MCP_PORTis not set, the server uses stdio transport (default behavior). MCP_HOSTdefaults to127.0.0.1(localhost only). Use0.0.0.0to allow connections from any network interface.- The HTTP/SSE transport is useful for remote access, but stdio is recommended for local IDE integrations.
- Each HTTP request can have its own Authorization header, allowing multi-tenant use cases.
For self-hosted GitGuardian instances, add the GITGUARDIAN_URL environment variable to your MCP configuration:
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": ["--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server"],
"env": {
"GITGUARDIAN_URL": "https://dashboard.gitguardian.mycorp.local"
}
}
}
}If your self-hosted instance has honeytokens enabled and your user has the required permissions ("manager" role), you can explicitly request honeytoken scopes:
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": ["--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server"],
"env": {
"GITGUARDIAN_URL": "https://dashboard.gitguardian.mycorp.local",
"GITGUARDIAN_SCOPES": "scan,incidents:read,sources:read,honeytokens:read,honeytokens:write"
}
}
}
}For the GitGuardian EU instance, use:
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": ["--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server"],
"env": {
"GITGUARDIAN_URL": "https://dashboard.eu1.gitguardian.com"
}
}
}
}If you have your own OAuth application configured in GitGuardian, you can specify a custom client ID:
{
"mcpServers": {
"GitGuardianDeveloper": {
"command": "uvx",
"args": ["--from", "git+https://github.com/GitGuardian/ggmcp.git", "developer-mcp-server"],
"env": {
"GITGUARDIAN_CLIENT_ID": "my-custom-oauth-client"
}
}
}
}If you want to contribute to this project or add new tools, please see the Development Guide.
This project includes a comprehensive test suite to ensure functionality and prevent regressions.
-
Install development dependencies:
uv sync --dev
-
Run the test suite:
ENABLE_LOCAL_OAUTH=false uv run pytest
Note: Tests disable OAuth by default via the
ENABLE_LOCAL_OAUTH=falseenvironment variable to prevent OAuth prompts during test execution. -
Run tests with verbose output:
ENABLE_LOCAL_OAUTH=false uv run pytest -v
-
Run tests with coverage:
ENABLE_LOCAL_OAUTH=false uv run pytest --cov=packages --cov-report=html
This will run all tests and generate a coverage report showing which parts of the codebase are covered by tests.