Skip to content
638Labs
Docs Blog

Quickstart

Get started with the 638Labs API in three steps: get a key, call a native agent, then register your own endpoint.

1. Get an API Key

  1. Visit 638Labs.com and create an account.

  2. Go to the API Keys section in your dashboard sidebar. Create an API key or use the default one created with your account.

  3. Copy the API key to your clipboard (reveal key in the right-hand action menu to copy).

  4. Set the environment variable in your terminal:

Terminal window
export STOLABS_API_KEY=<your-api-key>

We use STO as shorthand for 638 - API variable names can’t start with numbers.

2. Call a Native Agent

638Labs has native agents you can call immediately. Let’s try stolabs/BulletBot, a summarization agent:

Terminal window
curl -s https://sto0.638labs.com/api/v1 \
  -H "Content-Type: application/json" \
  -H "X-Stolabs-Api-Key: $STOLABS_API_KEY" \
  -H "X-Stolabs-Route-Name: stolabs/BulletBot" \
  -d '{
    "messages": [
      {
        "role": "user",
        "content": "Summarize: Large language models are trained on vast amounts of text data. They learn patterns in language that allow them to generate coherent text, answer questions, and perform a variety of natural language tasks. These models have become foundational components in modern AI applications."
      }
    ]
  }'

The response follows the OpenAI chat.completion format:

{
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "• LLMs train on large text datasets to learn language patterns\n• Capable of text generation, Q&A, and NLP tasks\n• Now foundational to modern AI applications"
      }
    }
  ]
}

Try other native agents by changing the X-Stolabs-Route-Name header - for example, stolabs/TLDRBot for a 2-sentence summary, or stolabs/QuickChat for a conversational response. See the full list at Native Agents.

3. Register Your Own Endpoint

Now route a request through 638Labs to your own AI provider - for example, OpenAI.

Claim your route namespace

Every endpoint you register lives under a namespace prefix unique to your account (e.g. myorg/). You need to claim this once before creating endpoints.

  1. In the dashboard, go to Profile (bottom of the sidebar).
  2. In the Route Namespace card, enter your desired prefix (lowercase, letters/numbers/hyphens, 2-30 characters).
  3. Click Claim. This is permanent - choose carefully.

Your prefix becomes the first part of every route name you create: myorg/agent-name.

Create a registry entry

  1. Click Endpoints in the sidebar.
  2. Click New Endpoint.
  3. Your namespace prefix is pre-filled. Enter the endpoint name and details:
Route Name:  myorg/openai-gpt4o   (prefix is automatic)
URL:         https://api.openai.com/v1/chat/completions
Type:        AI Model

Route names follow a prefix/name format (e.g. myorg/openai-gpt4o).

Test it

Set your OpenAI key alongside your 638Labs key:

Terminal window
export OPENAI_API_KEY=<your-openai-key>

Then call your registered endpoint through the gateway:

Terminal window
curl -s https://sto0.638labs.com/api/v1 \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "X-Stolabs-Api-Key: $STOLABS_API_KEY" \
  -H "X-Stolabs-Route-Name: myorg/openai-gpt4o" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": "Write a one-sentence bedtime story about a unicorn."
      }
    ]
  }'

The gateway routes your request to OpenAI through your registered endpoint. Your 638Labs API key handles auth with the gateway; your OpenAI key is passed through to OpenAI via the Authorization header.

Next Steps

  • Browse Native Agents to see all 18 agents available for testing
  • Read the AI Registry docs for the full registration walkthrough with screenshots
  • Check the AI Gateway docs for authentication patterns and routing details
  • See Rate Limits for free and paid tier limits