GitGuardian MCP Server – Secrets Security
GitGuardian's official MCP Server brings secrets security into AI coding assistants. Use it to scan code for leaked credentials, inspect and remediate incidents, generate honeytokens, and work with GitGuardian security context from an IDE, chat client, CI workflow, or self-hosted environment.
Overview
GitGuardian's official MCP Server connects AI assistants to GitGuardian's
secrets security platform. It is designed to help developers and security teams
detect leaked credentials, understand incidents, create honeytokens, and
perform guided remediation without leaving an MCP-compatible IDE or chat client.
What the MCP server enables
The server exposes tools based on the OAuth scopes or personal access token
scopes granted to the connection. Depending on those permissions, an AI agent
can:
- Scan files or projects for leaked credentials using GitGuardian secret
detection. - List, filter, assign, resolve, and tag incidents.
- Work with both internal monitoring and Public Monitoring incidents.
- Retrieve incident details and remediation context.
- Generate honeytokens and list existing honeytokens.
- Open pull requests that remediate secrets in repositories monitored by the
GitGuardian workspace. - Support prompts such as scanning a codebase, checking incidents assigned to a
user, explaining a security incident, or creating a decoy credential for
monitoring.
When to use it
Use GitGuardian MCP when an AI agent is helping with secure development or
secrets incident response. Practical examples include scanning AI-generated code
before committing, triaging newly assigned incidents, preparing remediation
steps for a leaked token, creating a honeytoken for a sensitive repository, or
automating a pull request that removes an exposed credential from a monitored
project.
Connection and authentication
The recommended SaaS setup is hosted Streamable HTTP. Use
https://mcp.gitguardian.com/mcp for US SaaS and
https://mcp.eu1.gitguardian.com/mcp for EU SaaS. In hosted mode, the MCP
client handles OAuth automatically through GitGuardian's OAuth proxy flow.
Local stdio is supported for CI/CD, scripts, one-off invocations, airgapped
environments, or older clients. The documented command uses uvx --from git+https://github.com/GitGuardian/ggmcp.git gg-mcp-server with
ENABLE_LOCAL_OAUTH=false, GITGUARDIAN_PERSONAL_ACCESS_TOKEN, and
GITGUARDIAN_URL.
Self-hosted deployments can run the published Docker image
ghcr.io/gitguardian/ggmcp:latest behind a TLS-terminating reverse proxy. The
container exposes Streamable HTTP on port 8000 by default and can use OAuth
proxy mode for self-hosted GitGuardian dashboards.
Key considerations
GitGuardian MCP is documented as beta. MCP agents act on the user's behalf, so
review actions that affect incidents, honeytokens, or remediation pull requests.
The exact tools exposed depend on the scopes of the OAuth token or personal
access token. Hosted requests transit the GitGuardian MCP service in addition
to the GitGuardian API. Browser-driven OAuth in stdio mode is deprecated; new
stdio deployments should use a personal access token, while OAuth-driven flows
should use hosted or self-hosted HTTP. Store tokens securely and use the
smallest set of scopes needed for the workflow.
Supported Transports
streamable_http
URL: https://mcp.gitguardian.com/mcp
streamable_http
URL: https://mcp.eu1.gitguardian.com/mcp
stdio
Command: uvx
Args:
--fromgit+https://github.com/GitGuardian/ggmcp.gitgg-mcp-server
streamable_http
URL: https://<your-gitguardian-mcp-domain>/mcp
Frequently Asked Questions
- When should an AI agent use the GitGuardian MCP server?
- Use it when an agent needs to scan code for leaked secrets, triage GitGuardian incidents, explain remediation steps, generate honeytokens, or create supervised remediation pull requests for repositories monitored by a GitGuardian workspace.
- What does the GitGuardian MCP server add to an AI agent's capabilities?
- It gives the agent direct access to GitGuardian secret detection, incident management, honeytoken, and remediation workflows through MCP tools, so the assistant can work with live security findings instead of relying only on generic security advice.
- What can an AI agent access or manage through GitGuardian MCP?
- Depending on granted scopes, the agent can scan files or projects, list and filter incidents, assign, resolve, or tag incidents, inspect internal and Public Monitoring findings, list or create honeytokens, and open remediation pull requests for monitored repositories.
- How is authentication configured for GitGuardian MCP?
- Hosted US and EU SaaS endpoints use OAuth proxy mode handled by the MCP client. Local stdio should use ENABLE_LOCAL_OAUTH=false with GITGUARDIAN_PERSONAL_ACCESS_TOKEN and GITGUARDIAN_URL. The tools available to the agent depend on the OAuth or personal-access-token scopes granted.
- Which transport should be used for GitGuardian MCP?
- Use hosted Streamable HTTP for GitGuardian SaaS: the US endpoint is https://mcp.gitguardian.com/mcp and the EU endpoint is https://mcp.eu1.gitguardian.com/mcp. Use local stdio with a personal access token for CI/CD, scripts, one-off local use, or older clients, and use self-hosted Streamable HTTP when running GitGuardian in self-hosted or airgapped environments.