Skip to main content
Early access. Cloud sessions and managed agent runs are not open to every account yet: we are letting accounts in gradually while we test. They run in a cloud sandbox operated by Mutagent, so there is nothing to host. A sandbox with nothing to do for 15 minutes stops; send the session a message and it wakes up, delivers your message and carries on in the same conversation.
Your cloud sessions call your models with your own API keys. Add your LLM providers to the workspace once, and every session in that workspace can use them. The keys are stored encrypted.

​
Before you start

  • The mutagent CLI is installed. See Installation.
  • You are signed in. In a terminal, run mutagent login. Without a browser, set MUTAGENT_API_KEY and run mutagent login --json. A coding agent helping a person runs mutagent login --browser --json and shows them the printed URL. See Sign in.
  • A workspace is selected. With more than one, choose it:

​
Mirror your local Helix setup

If Helix already works on your machine, copy its LLM providers:
The command shows what it will copy and asks before it sends anything. It then:
  • copies API keys stored as plain values in your local Helix setup (~/.mutagent/agent),
  • copies custom LLM providers and the models they declare,
  • sets the workspace default model if the workspace has none. It never replaces a default.
It does not copy:
  • OAuth or subscription logins. One exception: an OpenRouter permanent key saved in OAuth form is copied as a plain API key.
  • Keys stored as references, such as $VAR, ${VAR}, or !command. They are never evaluated.
  • Your local default model.
The plan lists each LLM provider to create or update, any endpoint that changes, and anything that blocks a copy. A copied key does not prove that the model works. To confirm it, run a task with that model. In a script, run mutagent providers mirror --json first. Without --yes it writes nothing, prints the plan, and exits with code 1. Review the plan, then run it again with --yes --json. Exit code 3 means you are not signed in or no workspace is set. Exit code 1 means at least one entry failed.

​
Add an LLM provider directly

Pipe the key in with --api-key-stdin, so it stays out of your shell history:
--api-key <api-key> also works, but it puts the key on the command line. Run mutagent providers list --json first to check whether the LLM provider already exists. A coding agent confirms with the user before it adds one: the command stores a credential. The key is tested against the LLM provider before it is saved. If the workspace has no default model yet, adding the LLM provider sets one. An existing default is never changed. See the command reference for every type and its flags.

​
Models and the workspace default

This lists every model your sessions can use, as provider/model, and marks the default with ★. A run picks its model in this order:
  1. --model <provider/model>, if you pass it. It must be in the list, or the run is refused with HTTP 422.
  2. The workspace default.
  3. Neither: the run is refused with HTTP 428 before any sandbox starts. The message says what to set up.
To choose the default yourself:
You can list fallbacks after the first ID. They are used in order when the LLM provider of an earlier model has been deactivated. The command replaces the whole list. --clear removes it, after which a run without --model is refused with HTTP 428. The cloud default and your local Helix default are separate. Changing one does not change the other.

​
Check that it worked

Setup is done when models is not empty and default is a model ID, not null. Pass an ID from models[].id to --model; never make one up.

​
If it fails

See Errors and exit codes for every code, and Troubleshooting cloud runs for the HTTP 422 and 428 refusals.