`run_workflow`

Orchestration write

What it does

Executes a workflow template, threading outputs between steps via dot-path references (input.X, previous.X, item.X). Returns a per-step log plus the final aggregated output.

Workflows respect a soft 25-second budget per Cloudflare request; long runs that hit the budget return status: "partial" with a resume cursor (cron continuation is partially shipped — see Limitations).

Input schema

Name	Type	Required	Default	Description
template	string	required	`—`	Template key from `list_workflow_templates`.
input	object	optional	`{}`	Top-level input bag — referenced from steps as `input.<key>`. Shape depends on the template.
dry_run	boolean	optional	`false`	Validate and resolve args without invoking tools. Returns the planned step list.

Example call

json{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "run_workflow",
    "arguments": {
        "template": "ads_pipeline",
        "input": { "video_url": "https://veo.example/cdn/v3.mp4", "title": "Coffee morning v3", "brand": "CorpoSostenibile" }
    }
  }
}

Response shape

json{
  "ok": true,
  "template": "ads_pipeline",
  "execution_id": "exec_4f2a...",
  "started_at": 1716301200,
  "completed_at": 1716301223,
  "duration_sec": 23,
  "status": "success",
  "abort_reason": null,
  "resume_cursor": null,
  "steps": [
    { "step": 1, "tool": "upload_video", "status": "success", "duration_ms": 5400, "result": { } },
    { "step": 2, "tool": "analyze_video", "status": "success", "duration_ms": 8900 },
    { "step": 3, "tool": "compliance_scan", "status": "success", "duration_ms": 4100 }
  ],
  "final_output": { "ok": true, "id": "c_abc123", "new_stato": "Attesa Approvazione" }
}

Usage from Claude Desktop

prompt"Run the ads_pipeline workflow on this Veo URL: https://veo.example/cdn/v3.mp4 — title 'Coffee morning v3', brand CorpoSostenibile."

Usage from Python

pythonimport requests

LUMIQA = "https://lumiqa.io/mcp/your-slug"
HEADERS = {"Authorization": "Bearer lk_live_xxxxxxxxxxxx"}

r = requests.post(LUMIQA, headers=HEADERS, json={
    "jsonrpc": "2.0", "id": 1,
    "method": "tools/call",
    "params": {
        "name": "run_workflow",
        "arguments": {"template": "ads_pipeline", "input": {"video_url": "https://...", "title": "v3"}}
    },
})
print(r.json())

Usage from Node

javascriptconst r = await fetch("https://lumiqa.io/mcp/your-slug", {
  method: "POST",
  headers: {
    "Authorization": "Bearer lk_live_xxxxxxxxxxxx",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    jsonrpc: "2.0", id: 1,
    method: "tools/call",
    params: {
      name: "run_workflow",
      arguments: { template: "ads_pipeline", input: { video_url: "https://...", title: "v3" } },
    },
  }),
});
const data = await r.json();
console.log(data);

Required env vars / integrations

Inherits the secrets of every tool the template calls (e.g. ANTHROPIC_API_KEY + OPENAI_API_KEY for ads_pipeline).
D1 table workflow_executions is written best-effort — missing table doesn't break a run.

Known limitations

25-second soft budget per request. Long workflows return status: "partial" with a resume_cursor; the cron continuation worker is shipped partially — budget-aware retries are on the v1.1 roadmap.
_for_each iterations are capped at 50 per run; bigger fan-outs need a paginated template.
on_error: "abort" (default) halts the workflow; continue swallows the error and moves on.
dry_run: true returns resolved args without invoking any tool — useful for plan inspection.

run_workflow