Start here

Set up your accounts.

Accounts tell CrazyToad which providers it may call for the brain, for agents, and for media. This guide walks each path field by field: subscription CLI logins, API keys, and local models. Open Manage → Accounts and the cheapest path is a first-party CLI login — you sign in once with the provider and CrazyToad drives that session, no key to paste.

At a glance

  • Accounts tell CrazyToad which providers it may call for the brain, for agents, and for media.
  • The cheapest path is a first-party CLI login — sign in once and CrazyToad drives the session, no key to paste.
  • Use API keys for billable providers (OpenAI-compatible, Grok/xAI, the Gemini API, media models).
  • Local models give you a no-cloud lane through Ollama, LM Studio, vLLM, or llama.cpp.
manage — accounts
CrazyToad account management page with provider cards and defaults
Note
The Accounts page: provider cards, defaults, health badges, and the add buttons.

Sign in with a subscription CLI

Go to Manage → Accounts and find the card titled First-party CLI logins. Click Claude (subscription) to run claude auth login, Codex (subscription) to run codex login, or Antigravity CLI (Google account) to run agy. A window opens and you finish the provider’s own sign-in there — it is an OS-level interactive login, not a token you paste. The new account appears as a card with a health badge; press Test model to confirm “CLI auth OK”. A subscription login lets the brain spawn real terminal sessions with no key to manage.

Note
Subscription Claude and Codex logins need no key. The Antigravity CLI signs in with your Google account but is agents-only — it cannot drive the brain, only the agent lane.

Add an API key — the 4-step wizard

Use a key for any provider that is not a local CLI subscription — an OpenAI-compatible endpoint, Grok/xAI, the Gemini API, or media models. Click + Add account (or a quick-add button) to open the 4-step wizard: 1 Provider · 2 Endpoint · 3 Billing & key · 4 Models & roles. Each step captures one part of the account.

LabelAny name you choose, e.g. “Claude personal subscription”. Required.
Provider kindWhich provider this account is, picked from the select after choosing a tile.
Endpoint URLMust be a valid URL. OpenAI https://api.openai.com/v1; Grok/xAI https://api.x.ai/v1.
Billing kindSubscription credits, Billable API, or Free. Only Billable API stores a key.
API keyPaste under New key, or reuse a vault key. Gemini keys start with AIza (Google AI Studio).
Default modelRequired. Free-text model id (or a curated Gemini dropdown). Press Refresh models to pull ids from the endpoint.
RolesToggle chips: Brain, Agent CLI, Image, Video, Other media. Brain/Agent CLI are required to set a default brain/agent.
Warning
Keys are stored encrypted with Windows DPAPI or the macOS Keychain. They live in a shared Stored keys vault (Label + key) that is reused for API keys and Remote Agents SSH keys.

Add a local model

Run models on your own machine through Ollama, LM Studio, vLLM, or llama.cpp. Click + Local Ollama (or the Local tile in the wizard) — billing is Free, no key is stored, and CrazyToad discovers what you have installed.

Endpoint flavorOllama, LM Studio, vLLM, or llama.cpp — match it to the runtime you are running.
Endpoint URLOllama http://localhost:11434/v1; LM Studio http://localhost:1234/v1.
Refresh modelsPulls installed model ids from the endpoint’s /models so you do not have to type them.
Default modelAn actually-installed id from the list, e.g. qwen2.5-coder:14b.
Warning
Don’t paste a Codex or Gemini CLI model id as the Ollama default — it will 404. Pick a model that is really installed in your local runtime.

Set defaults & route roles

Once accounts exist, tell CrazyToad which one to reach for by default and how to split work across roles. Each account card carries the default buttons; a separate Agent roles card maps roles to providers. Both auto-save, and each route is tuned by three fields.

ProviderWhich subscription drives the role: Claude, Codex, or Antigravity.
ModelOptional override for that role. Leave blank to use the provider’s default model.
EffortReasoning depth for the role: Minimal, Fast, Standard, Deep, or Max.
Note
Verify reachability before you rely on a route: press Test model on a card for a one-token check (“Model OK (NNN ms)” or “CLI auth OK”), use the wizard’s Test connection for Gemini, and watch the passive health badges. Installed is not the same as usable. All of this lives under Manage.