Use Autark in Zapier.

Add a step to any Zap that asks Autark a question and uses the answer in the steps that follow. Summarize an email, classify a lead, draft a reply, extract fields from a form: the Zap you already have gains a step that thinks.

How this works, and what it is not

Zapier's built-in AI steps let you bring your own key, but only from a fixed list of providers: OpenAI, Anthropic, Google Gemini, Azure OpenAI, and Amazon Bedrock. There is no field for a custom address, so Autark cannot be selected as the AI provider inside those steps.

What works instead is putting Autark in the middle of the Zap as a webhook step. You get the same result, and it works in every Zapier plan that includes Webhooks by Zapier. It is a few more clicks than picking a provider from a dropdown, and it is the only way to run Autark inside a Zap today.

Add the step

Place this step wherever you want the AI work to happen. Everything before it can feed it information; everything after it can use the answer.

1. Add a step after your trigger
2. App:      Webhooks by Zapier
3. Event:    Custom Request
4. Method:   POST
5. URL:      https://api.autark.ai/v1/chat/completions
6. Payload Type: json
7. Data:     paste the body below (see the next block)
8. Headers:  Authorization  ->  Bearer autark_sk_YOUR_KEY
             Content-Type   ->  application/json
9. Test the step

// Custom Request is the event that lets you set your own URL and headers.

The request body

This is the part that looks technical. You only need to change two things: the sentence you want answered, and which Autark model answers it. Pull in data from earlier steps with Zapier's field picker where you see {{message_text}}.

{
  "model": "autark-agent",
  "messages": [
    {
      "role": "user",
      "content": "Write a one-line summary of this message: {{message_text}}"
    }
  ]
}

// model: autark-agent for everyday work, autark-pro when answers feel shallow.

Using the answer in later steps

Zapier parses the response automatically. The text Autark wrote appears as a field you can map into any following step.

{
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "The summary Autark wrote."
      }
    }
  ]
}

// In later Zap steps, map the answer with:
//   Choices Message Content

// Pick the field from the step's output list rather than typing it.

Which model to use

Start with autark-agent. It handles classification, summarization, extraction and drafting, which is what most Zaps need, at the lowest cost. Move to autark-pro when the answers feel shallow, and autark-deep only for genuinely hard reasoning. Add -private when the content is sensitive: same quality, restricted to providers that never retain or train on your prompts.

Zaps run on a schedule, so a step that fires often adds up. Autark picks the model per request, which is what keeps a high-volume Zap affordable.

Troubleshooting

The step fails with 401 or 403

The Authorization header is missing or malformed. It must read Bearer autark_sk_YOUR_KEY, including the word Bearer and the space after it.

The step fails with 404

The URL is missing /v1 or the path after it. It should be https://api.autark.ai/v1/chat/completions.

The step succeeds but the answer is empty

Check the model name is spelled exactly as listed, including any -private suffix. A wrong model name returns an error rather than an empty answer, so an empty answer usually means the field mapping in a later step is pointing at the wrong output.

Can I use this in a Zapier Chatbot or Agent?

Not as the model behind it. Those products select their own models and do not accept a custom endpoint. The webhook step above works in a Zap, and a Zap can be triggered by a chatbot conversation.

All integrations API reference