# Call a Task

Tasks are served through an OpenAI-compatible API, so any OpenAI SDK or client works. Set the base URL to your serving URL (shown in each Task's **Use it** panel) and the model to the Task's name.

```typescript
import OpenAI from "openai";

const client = new OpenAI({ baseURL: "<your serving URL>", apiKey: process.env.FACET_API_KEY });
const reply = await client.chat.completions.create({
  model: "support-triage",
  messages: [{ role: "user", content: "My card was charged twice" }],
});
```

```sh
curl <your serving URL>/chat/completions \
  -H "Authorization: Bearer $FACET_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "support-triage", "messages": [{"role": "user", "content": "My card was charged twice"}]}'
```

## Versions

- `model: "support-triage"` follows Latest: when a Change merges, callers get it at once.
- `model: "support-triage@v4"` pins Version 4.

The response's `x-facet-version` header names the Version that answered, and `x-facet-request-id` identifies the call.

## Keys

API keys start with `fct_`. Make them under **Workspace settings, Keys**; a key can be limited to one Task, and a viewer key only reads. Keep keys in your app's secrets, never in code. Facet keeps only each call's metadata (Version, key, tokens, latency), never the messages or answers.

## Instructions

Facet adds the Task's Instructions to every call, so send only the input. Don't add your own system prompt: Facet owns the Instructions and improves them with each Change.