> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pre.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# From a pre.dev app

> Projects built on pre.dev have the gateway key pre-set; ask the build agent to add an AI feature.

Every project built on pre.dev gets the workspace's API key injected into its sandbox and deploy environment automatically. There is nothing to paste and nothing to rotate by hand.

| Variable         | Value                      |
| ---------------- | -------------------------- |
| `PREDEV_API_KEY` | The workspace key, `pdk_…` |
| `PREDEV_API_URL` | `https://api.pre.dev`      |

Build the gateway URL as `${PREDEV_API_URL}/v1`.

## Ask the build agent

Tell the coding agent what you want, for example **"add an AI feature"**, "add a summarize button that uses a fast model", or "let users chat with their documents". The agent reads the model catalog for you, picks an id, and wires a server-side call using the variables above. Ask it to list models with prices if you want to choose.

Everything the agent adds is ordinary code you can read and change. The credits a deployed feature spends come from the same workspace balance; turn on auto-recharge in billing so a live app does not stop at zero. See [pricing](/ai-gateway/pricing).

## Server-side only

The key spends the workspace's credits, so it must never reach the browser. A single-page app calls its own backend route, and the route calls the gateway. The build agent follows this rule; keep it when you edit by hand.

<CodeGroup>
  ```typescript Node.js server route theme={null}
  import OpenAI from 'openai';

  const client = new OpenAI({
    baseURL: `${process.env.PREDEV_API_URL}/v1`,
    apiKey: process.env.PREDEV_API_KEY!,
    defaultHeaders: { 'x-predev-project-id': process.env.PREDEV_PROJECT_ID ?? '' },
  });

  // POST /api/summarize  { text: string }
  export async function summarize(text: string) {
    const completion = await client.chat.completions.create({
      model: 'deepseek/deepseek-v4.1-flash',
      messages: [
        { role: 'system', content: 'Summarize the user text in three bullets.' },
        { role: 'user', content: text },
      ],
    });
    return completion.choices[0].message.content;
  }
  ```

  ```python Python server route theme={null}
  import os
  from openai import OpenAI

  client = OpenAI(
      base_url=f"{os.environ['PREDEV_API_URL']}/v1",
      api_key=os.environ["PREDEV_API_KEY"],
      default_headers={"x-predev-project-id": os.environ.get("PREDEV_PROJECT_ID", "")},
  )


  # POST /api/summarize  {"text": "..."}
  def summarize(text: str) -> str:
      completion = client.chat.completions.create(
          model="deepseek/deepseek-v4.1-flash",
          messages=[
              {"role": "system", "content": "Summarize the user text in three bullets."},
              {"role": "user", "content": text},
          ],
      )
      return completion.choices[0].message.content
  ```
</CodeGroup>

The browser calls `POST /api/summarize`; only the server holds the key. Validate and rate-limit that route as you would any endpoint that spends money.

## Attribute usage to the project

Send `x-predev-project-id: <projectId>` on each gateway call, as the examples above do, and the call is attributed to that project. Only the workspace's own projects count; other ids are ignored. [`GET /v1/usage`](/ai-gateway/api/usage) reports the attributed totals.

## Local development

Outside the sandbox, copy the key from **Integrations → API Keys** into your local environment as `PREDEV_API_KEY` and set `PREDEV_API_URL=https://api.pre.dev`. See [API key](/ai-gateway/api-key) and [environment variables](/coding-agent/integrations/env-vars).