Agent webhook triggers

An agent's Workers tab exposes an HTTP endpoint another system can POST to. The POST runs the agent against the payload and returns the reply; nothing in the UI is involved. Reach for it when something outside Tale needs the agent to answer a question — a Slack bot, a form handler, a scheduled job.

This page covers the Workers surface only. For the developer-facing equivalent (calling Tale from arbitrary scripts), see Develop → API reference; for inbound automation triggers that run a workflow rather than an agent, see Automations → triggers.

A worked Worker

Open the agent and switch to Workers. The page shows a per-agent URL and a sample curl. POST a JSON payload with a message field; Tale receives it, runs the agent against the message, and returns the agent's reply as the response body. The same payload sent twice produces two independent runs — Workers do not de-duplicate.

Authentication

Worker endpoints require an API key. The endpoint shows the URL but not a working curl until a key is attached; the Authorization header carries the key as a bearer token. Rotating the key invalidates every running caller — provision new keys before retiring old ones. API keys are managed under API keys.

Payload shape

The default payload is {"message": "…"}. Additional fields the agent's instructions reference can be added; they pass through to the model context as structured input. The agent's reply is returned as a JSON object with the reply text, any tool calls, and any citations. Streaming is supported when the caller sets the Accept: text/event-stream header.

Where this fits

Workers are the lightweight, per-agent equivalent of the API. They are useful when the integration is "this one agent does this one thing"; for richer flows, model the call as an Automation and point the integration at the automation's webhook trigger. The Trigger automation via webhook tutorial walks the automation shape end to end.