MCP Trigger

Expose this workflow as an MCP tool callable by external AI agents (Claude Desktop, Cursor, Windsurf, etc.).

When to Use

Use MCP Trigger when:

You want external AI agents to discover and call your workflow as a tool
You need to expose business logic (search, create, lookup) to MCP-compatible clients
Other agents should invoke your workflow and receive a structured response via a Return node

For connecting your workflow's AI Agent to external MCP servers, use MCP Tool Provider instead.

Output

Handle	Type	Description
`mcp`	object	MCP tool call context

The mcp (accessed as {{ mcp.* }}) shape:

Field	Type	Description
`tool_name`	string	The tool name from the MCP call
`arguments`	any	Full arguments object passed by the calling agent
`call_id`	string	Unique identifier for this tool invocation

Settings

Tool Definition

Setting	Type	Default	Description
`tool_name`	string	—	The name external agents see when discovering this tool (e.g. `create_invoice`, `search_docs`). Required.
`tool_description`	text	—	Help agents understand when and how to use this tool. Be descriptive. Required.
`parameters`	code editor (JSON)	JSON Schema with `query` property	JSON Schema defining what parameters the tool accepts. External agents validate arguments against this schema. Required.

Behavior

Setting	Type	Default	Description
`responseMode`	select	`use-workflow-response`	`use-workflow-response` waits for the workflow Return node and sends the result back to the agent. `ack-immediately` returns a success message instantly.
`timeout`	number	`30`	Maximum seconds to wait for workflow response before timing out (only applies in `use-workflow-response` mode).

Testing

Setting	Type	Default	Description
`mockPayload`	code editor (JSON)	—	JSON body to inject when running this workflow locally for testing.

Example

Canvas

A documentation search tool exposed to MCP clients:

Set tool_name to search_docs
Set tool_description to "Search the documentation knowledge base and return relevant results"
Set parameters to a JSON Schema with query (string, required) and limit (number, optional)
Connect the mcp output to your search logic, then wire a Return node

Access tool call data in downstream nodes:

{{ mcp.arguments.query }}
{{ mcp.arguments.limit }}
{{ mcp.call_id }}

After deploying, add the tool to your MCP client config:

{
  "mcpServers": {
    "tensorify-tools": {
      "url": "https://mcp.tensorify.io/v1/triggers",
      "headers": {
        "Authorization": "Bearer tfk_your_api_key_here"
      }
    }
  }
}

Or run tensorify mcp config to generate the config automatically.

TSL

import mcp from @tensorify/mcp-trigger:1.0.0
import http_request from @tensorify/http-request:3.0.0
import return_node from @tensorify/return:3.0.0

node trigger @tensorify/mcp-trigger:1.0.0 {
    tool_name = "search_docs"
    tool_description = "Search the documentation knowledge base"
    parameters = "{\"type\": \"object\", \"properties\": {\"query\": {\"type\": \"string\"}}, \"required\": [\"query\"]}"
    responseMode = "use-workflow-response"
    timeout = 30
}

node search @tensorify/http-request:3.0.0 {
    method = "GET"
    url = "https://api.example.com/search?q={{ mcp.arguments.query }}"
}

node respond @tensorify/return:3.0.0 {
    returnMode = "simple"
}

trigger.mcp -> search.body
search.response -> respond.input

Common Gotchas

When responseMode is use-workflow-response, you must connect a Return node — otherwise the agent's call times out after timeout seconds.
tool_name, tool_description, and parameters are all required. Agents use the description and schema to decide when and how to call your tool.
parameters must be valid JSON Schema. Invalid schema prevents agents from calling the tool correctly.
MCP Trigger exposes your workflow as a tool (server role). MCP Tool Provider connects your AI Agent to external tools (client role).