Workflow Tools

MCP tools for running durable, multi-step workflows. Workflows chain multiple tools (skills, connectors) into pipelines with durable, server-persisted state. A run can suspend to wait for human input and resume when answered.

start-workflow launches a workflow by ID and returns its runId; search-workflows finds the ID; list-schedules, set-schedule and unset-schedule manage a workflow’s schedule. Managing runs afterward (list, inspect, cancel, resume) is done through the REST API — see Workflows API.

For a walkthrough, see the Workflows Guide.

Finding workflow IDs

Search for a workflow over MCP with search-workflows (by name, description or capability), or list them over REST, to get a valid workflowId:

GET /api/v1/workflows

start-workflow

Start a workflow by ID. The tool launches the run and returns its runId; in chat the call needs your approval first.

Parameters:

Parameter	Type	Required	Description
`workflow_id`	string	Yes	Workflow ID to start
`workflow_name`	string	No	Human-readable name for display
`inputs`	object	No	Input parameters for the workflow

Response:

{
  "success": true,
  "data": {
    "runId": "wf_abc123",
    "workflowId": "deep-research",
    "conversationId": null
  }
}

Managing runs (REST)

Run lifecycle isn’t exposed as MCP tools; use the REST endpoints:

Action	Endpoint
List runs	`GET /api/v1/workflows/runs`
Get run details + step history	`GET /api/v1/workflows/runs/:runId`
Cancel a run	`POST /api/v1/workflows/runs/:runId/cancel`
Resume a suspended run	`POST /api/v1/workflows/runs/:runId/resume`

See Workflows API for request/response shapes.

Workflow States

Status	Description
`running`	Actively executing steps
`suspended`	Paused awaiting human input (resume to continue)
`completed`	All steps finished successfully
`failed`	Execution failed
`cancelled`	Cancelled by the user