Endor Labs MCP Server – AI Code Security
Endor Labs' official MCP server brings application security checks into AI coding workflows. Use it to check dependency safety, scan open source dependencies, detect leaked secrets, retrieve vulnerability context, and run Enterprise security reviews on code changes from an MCP-compatible IDE or CLI.
Overview
Endor Labs MCP Server is the official Model Context Protocol integration for
bringing Endor Labs security intelligence into AI-powered coding tools. It runs
locally as a lightweight process on the developer's machine, and compatible IDEs
or CLIs launch it automatically over stdio when an AI agent needs security
context.
What the MCP server enables
Endor Labs documents MCP tools for dependency and code security workflows. An
AI agent can:
- Check whether a dependency in a project is vulnerable.
- Check a dependency for broader security risks, including vulnerabilities and
malware. - Retrieve details for a vulnerability from the Endor Labs vulnerability
database. - Retrieve context from commonly used Endor Labs resources such as findings,
vulnerabilities, and projects. - Run an Endor Labs security scan to detect risks in open source dependencies,
common security issues, and credentials accidentally exposed in a Git
repository. - Run
security_reviewon code diffs in Enterprise Edition. This analyzes
local uncommitted staged and unstaged changes compared toHEAD, or diffs
between the main branch and the last commit.
When to use it
Use Endor Labs MCP when an AI coding agent is adding dependencies, modifying
manifests, generating code, or preparing a pull request. Practical workflows
include checking a package before installation, scanning a repository for leaked
secrets before commit, reviewing generated code for security issues, retrieving
vulnerability details for remediation, and giving an agent guardrails before
code review or CI surfaces the same issue later.
Connection and authentication
The default local stdio configuration runs npx -y endorctl ai-tools mcp-server. Node.js 18 or later is required for npx, and Endor Labs
recommends Node.js 24 LTS or later. If npx times out behind a proxy or a
system-installed endorctl is preferred, the MCP client can run endorctl ai-tools mcp-server directly.
Enterprise workflows may require a namespace and platform configuration.
Endor Labs documents Enterprise authentication settings such as
ENDOR_MCP_SERVER_AUTH_MODE and ENDOR_MCP_SERVER_AUTH_TENANT when SSO or
tenant-specific authentication is needed.
Key considerations
The MCP server performs security-sensitive analysis inside the developer
workflow, so run it only in trusted IDEs and repositories. Enterprise
security_review requires Enterprise Edition, a configured namespace, AI
security code review enabled in the Endor Labs platform, and appropriate user
permissions. Tool availability can be limited by client configuration because
Endor Labs allows disabling specific tools after setup. Store authentication
settings and tokens outside source control, review generated fixes, and confirm
remediations with tests, scans, and normal code review before merging.
Supported Transports
stdio
Command: npx
Args:
-yendorctlai-toolsmcp-server
stdio
Command: endorctl
Args:
ai-toolsmcp-server
Frequently Asked Questions
- When should an AI agent use the Endor Labs MCP server?
- Use it when an AI coding agent is adding dependencies, changing manifest or lock files, generating code, preparing a pull request, checking for leaked secrets, or remediating vulnerabilities with Endor Labs security context.
- What does the Endor Labs MCP server add to an AI agent's capabilities?
- It gives the agent MCP tools for dependency vulnerability checks, dependency risk and malware checks, vulnerability lookup, Endor Labs resource retrieval, repository security scans, secret detection, and Enterprise code-diff security review.
- What can an AI agent access or manage through Endor Labs MCP?
- The agent can check dependency safety, retrieve vulnerability details, access relevant Endor Labs findings and project context, run security scans for open source risks and exposed credentials, and run Enterprise security_review on code diffs when the namespace and platform settings allow it.
- How is authentication configured for Endor Labs MCP?
- The server runs through endorctl and uses the local Endor Labs authentication context. Enterprise setups may require ENDOR_NAMESPACE and settings such as ENDOR_MCP_SERVER_AUTH_MODE and ENDOR_MCP_SERVER_AUTH_TENANT. Store tokens and tenant values outside source control.
- Which transport should be used for Endor Labs MCP?
- Use local stdio. The documented default is `npx -y endorctl ai-tools mcp-server`. Use `endorctl ai-tools mcp-server` directly when endorctl is already installed or when npx is unsuitable behind a corporate proxy.