MCP Client

Call a tool on a remote MCP server as a standalone agent step

Written By pvdyck

Last updated 1 day ago

MCP Client

✅ Supported — Copied from n8n with platform patches

| Auth | None, Bearer Auth, Header Auth, Multiple Headers Auth, MCP OAuth2 |

Compatibility

FeatureStatus
HTTP Streamable transportSupported
SSE transportNot supported — blocked at validation
Tool listing (tools/list)Supported
Tool execution (tools/call)Supported
Image/audio binary responsesSupported

Version Requirements

  • v1 supported — v1 defaults to HTTP Streamable transport
  • If an agent pins serverTransport to sse, upload validation rejects it with a clear message. Switch to HTTP Streamable and re-upload.

Usage

Use this node as a standalone transform in your agent — it takes input items, calls a single tool on a remote MCP server, and returns the result. Unlike MCP Client Tool, which is a sub-node that exposes a whole MCP server to an AI Agent, MCP Client calls one specific tool directly and returns its output so downstream nodes can branch on it.