5-minute quickstart

One goal: get a first reply from an agent in a fresh environment.

Before you start

• An account (provisioned by the platform)
• An API key and model ID from a model provider

Step 1 — Sign in

Open the platform URL and sign in. You land in the workspace: the left side is the menu, the center is the session area.

Step 2 — Configure a model

1. Open LLM in the left menu.
2. Click New LLM config.
3. Fill in:
   • Display name — something you will recognise, e.g. “production”.
   • Protocol — must match your provider.
   • Model ID — the provider’s model identifier.
   • API key — paste the full key.
4. Save. Optionally set it as the default model so new sessions use it automatically.

You can leave API base URL, proxy and context window empty; the protocol defaults are usually enough.

Step 3 — Create a session

1. Click New session in the session area.
2. Pick an agent if you have one; otherwise use the default configuration.
3. Optionally name the session, then create it.

Step 4 — Send your first message

Type something like:

Describe what you can do in three sentences.

The reply streams in as it is generated.

Acceptance checklist

• The new model appears in the LLM list
• A new session can be created
• A message produces a complete reply
• The session is not marked as failed afterwards

If it does not work

Symptom	Check first
401 / auth failure	Key is complete and valid; protocol matches the provider
Model not found	Model ID spelling and account permissions
No reply at all	Network access to the API base URL; try another model
Reply stops mid-way	Any error shown in the session; retry settings under Setting → Basic

More in the FAQ.