nodes / builtin / ai / aiagents
Run agent
Give a task to one of your agents and get its answer.
Give a task to one of your agents and get its answer.
What it does
Section titled “What it does”Run agent hands a task, written in plain language, to one of your agents or to Aria. The agent works on it like a chat turn in the background, using its own model, instructions, memory, skills and rules, and the node returns its reply.
This is the node a routine runs: a trigger followed by Run agent. Because a routine is an ordinary workflow, you can open it in the canvas and add steps before or after the agent.
Nobody is there to approve actions during a background run. Actions the agent’s rules mark as “ask” are skipped and listed in Activity instead.
When to use it
Section titled “When to use it”- Run an agent on a schedule: “Every morning, check my three saved trips for price drops and tell me what changed.”
- Let an agent react to a page: put a Page Load trigger in front of it.
- Hand off a judgement call inside a longer workflow, then use the reply in the next steps.
To give an agent extra abilities inside one workflow (tools, a specific model), use the Tools Agent node instead.
Inputs and settings
Section titled “Inputs and settings”| Setting | Type | What it does |
|---|---|---|
| Agent | Text | The agent (id or @handle) that does the task. Default: Aria. Default: aria |
| Task (required) | Long text | What the agent should do, in plain language. You can use expressions. |
| Save the result as a chat | Toggle | Keep the answer in your chats so you can read and continue it. Default: On |
Agent takes an agent id or its @handle, for example @travel. Leave it as aria to use Aria. Task accepts expressions, so you can pass in data from earlier nodes.
Outputs
Section titled “Outputs”-
Output
Each output item has these fields:
| Field | Type | What it holds |
|---|---|---|
| reply | Text | |
| agentId | Text | |
| agentName | Text | |
| conversationId | Text |
{ "reply": "…", "agentId": "…", "agentName": "…", "conversationId": "…"}reply is the agent’s answer, agentId and agentName name the agent that did the task, and conversationId points to the chat the result was saved to (only when Save the result as a chat is on).
Dependencies and credentials
Section titled “Dependencies and credentials”None. This node needs no connection and has no slots for other nodes.
The agent uses the model and connections set on the agent itself, so this node needs nothing extra.
Example workflow
Section titled “Example workflow”- A Schedule trigger runs every weekday at 8:00.
- Run agent with Agent
@newsand Task “Summarise the three most important stories on my saved sites since yesterday.” - A Telegram node sends
{{ $input.reply }}to your phone.
Common issues
Section titled “Common issues”- The agent says it couldn’t do part of the task. An action needed approval and was skipped. Check Activity, then change the rule to allow that action in routines, or run the task in a chat instead.
- The wrong agent answered. Check the id or @handle in Agent. An empty value falls back to Aria.
- No chat appears. Turn on Save the result as a chat.
Related nodes
Section titled “Related nodes”- Tools Agent: an agent built inside the workflow, with its own model and tools.
- Chat Trigger: start a workflow from a chat message.
- Schedule: run a routine at set times.