Skip to content
ATLAS
 Docs
On this page

Quickstart#

In about five minutes, you'll sign in, create an API key and make your first model call through the ATLAS gateway.

Before you begin#

  • Your organisation is set up on ATLAS and you have an account in it. If not, sign-in ends on an error page such as Access Refused or Organization Not Recognized. Ask your organisation's Admin or your ilionx contact.
  • You are a member of at least one team space or active project that has a model enabled. A key acts under a space or project, its acting context. The acting context decides which models you can call and whose budget pays.
  • You know your ATLAS address. Replace https://<atlas-host> with your ATLAS address. The web app and the gateway share it, and every gateway URL starts with /gateway.

1. Sign in#

  1. Open the web app and select Log in.
  2. Enter your work email address. ATLAS recognises your organisation from the email domain and sends you to your company's login page.
  3. Sign in. You land in the app, and the Personal group in the sidebar holds Chat, My usage, My conversations and My API keys.

2. Create an API key#

  1. Open My API keys and select Create API key.
  2. Choose a Context, the space or project the key acts under.
  3. Enter a Name, for example quickstart.
  4. Choose an Expiration. 90 days is preselected, and your space may cap the maximum.
  5. Select Create Key.
  6. In the API Key Created dialog, copy the key. It starts with atl_sk_. The dialog says you won't see the key again. Until the key expires, you can still copy it with Copy Key in its row menu.

Store the key and your ATLAS address in environment variables:

bash
# BASE_URL is your ATLAS address, without a trailing slash
export BASE_URL="https://<atlas-host>"
export API_KEY="atl_sk_..."

3. Make your first call#

  1. List the models your key can use:

    bash
    curl -s "$BASE_URL/gateway/openai/v1/models" \
  -H "Authorization: Bearer $API_KEY"

    Each model id has the form provider-slug/model-slug, for example openai/gpt-4.1.

  2. Copy one id and send a chat completion:

    bash
    curl -s "$BASE_URL/gateway/openai/v1/chat/completions" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-4.1","messages":[{"role":"user","content":"Say hello"}]}'

    The response is a JSON completion with the model's reply in choices[0].message.content. The request appears in My usage and counts against the budget of your key's acting context.

Anthropic clients use $BASE_URL/gateway/anthropic as their base address. They send the same key as a bearer token or in an x-api-key header. For more information, see connect OpenAI and Anthropic clients.

4. Try the chat (optional)#

Open Chat and send a message from one of the starters (Summarise, Write code, Explain or Draft an email). The composer asks you to pick an API key, then a model, before it accepts a message.

If something fails#

Response What it means
401 authentication_required The key is missing, mistyped, expired or revoked.
404 model_not_found Use provider-slug/model-slug exactly as /models lists it. A bare name such as gpt-4.1 fails.
403 no_acting_context The key has no space or project. Use Switch Context in the key's row menu.

For every code, see Errors. For symptoms and fixes, see Troubleshooting.

Next steps#