Skip to main content

​
Overview

Tracecat supports two MCP integration types:
  • Remote MCP over URL (HTTP or SSE)
  • Local MCP over stdio

​
Availability

Open source In open-source deployments, bring your own remote or stdio MCP server. Select the saved integration in the mcp_integrations input on ai.agent. Enterprise Edition Enterprise includes a catalog of 50+ preconfigured MCP servers with guided connection setup, including Splunk, SentinelOne Purple AI, CrowdStrike Falcon, Microsoft Sentinel, Elastic, Wiz, GreyNoise, and PagerDuty. You can also attach MCP integrations to reusable agent presets. MCP options

​
Remote MCP

Use remote MCP when the server is exposed over a URL. Authentication modes:
  • No Authentication
  • Custom
  • OAuth
Remote MCP authentication

​
OAuth for remote MCP

For remote MCP with OAuth, link the MCP integration to an existing OAuth integration. Tracecat refreshes the token and sets the Authorization header automatically. For a custom remote MCP server, first create a custom OAuth provider in OAuth, then attach it to the MCP integration.

​
Custom headers for remote MCP

Custom authentication stores request headers as JSON. Header values support workspace secret and variable expressions.

​
Stdio MCP

Use stdio MCP when Tracecat should launch a local command such as npx, uvx, or a custom binary. stdio environment variables support Tracecat expressions.
Tracecat resolves those expressions from the workflow default environment unless overridden in action control flow.

​
Secrets and variables in MCP configuration

Both remote and stdio MCP integrations support template strings.
  • See AI agent for ai.agent and ai.preset_agent inputs and examples.
  • See OAuth for custom OAuth providers and OAuth token expressions.
  • See Secrets for environment-scoped secret resolution.
  • See Variables for non-secret values used in stdio environment variables.