Skip to main content
The Zero MCP server gives AI agents controlled access to the same vulnerability intelligence and operations available through the REST API. Use MCP when an agent needs to discover and call Zero operations dynamically. Use the REST API for application code that benefits from explicit HTTP requests and generated clients.

Connect an MCP client

Provision an API key and connect Codex, Claude Code, or another MCP client.

Endpoint

The server uses Streamable HTTP at:
For interactive clients, the MCP server uses OAuth 2.1 with PKCE. Codex and Claude Code discover the authorization flow automatically, open the Zero consent screen, and store a short-lived access token after you approve access to an organization. API keys remain available for non-interactive environments.

Tools

execute does not accept a URL, HTTP method, or arbitrary headers. It resolves those values from the live OpenAPI document, validates the supplied parameters and JSON body, and invokes only a catalogued operation.

Tool schemas

Review complete inputs, outputs, validation behavior, and example calls for both tools.

Search for an operation

The response includes operations such as search_vulnerabilities, including the full request schema. Use that schema to construct the next call.

Execute the operation

Zero-day history is vulnerability evidence, not a separate feed or entity. Query it alongside CVE, CVSS, EPSS, CISA KEV, affected software, and remediation fields through the vulnerability search operation.

API behavior

MCP uses the REST API’s entity schemas, query builder, pagination, authorization, and error behavior. Refer to the API documentation for those contracts instead of treating MCP as a separate data model. OAuth access tokens are bound specifically to the MCP resource. The MCP server validates them before processing JSON-RPC and uses separate internal service credentials when it calls the REST API.