Pick a provider

Anton works with any of these provider options. You choose one during onboarding (the first time you run anton), and you can switch at any time with /llm in chat or by running anton setup. All keys are persisted to ~/.anton/.env, so they carry across sessions and workspaces.

At a glance

Option	Default model	Web search / fetch	Best for
MindsHub — recommended	Resolved from the live catalogue (sonnet / haiku tier defaults)	Native passthrough, zero setup	Best overall experience
Anthropic (bring your own key)	claude-sonnet-4-6	Native server tools	Anthropic accounts
OpenAI (bring your own key)	gpt-5.4	Native via Responses API	OpenAI accounts
Google Gemini (bring your own key)	gemini-3-flash-preview	Needs anton setup-search	Gemini accounts
Custom OpenAI-compatible	You choose	Needs anton setup-search	Ollama, vLLM, Together, Groq, LM Studio, Azure, …

Option 1 — MindsHub (recommended)

MindsHub is the default and recommended choice. During setup Anton resolves the planning and coding models from the server's live model catalogue, picking the chat models your key is actually entitled to use (embedding and decision models such as Jev are skipped), and validates the connection with a real request against that configuration:

• Latest frontier models
• Faster responses
• Cost optimized
• Secure data connectors
• Native web search and fetch passthrough — no extra setup

During onboarding, if you don't have a MindsHub API key yet, Anton opens the signup page for you — it takes a few seconds.

The MindsHub host itself is not prompted for. To point Anton at a non-default host (staging, self-hosted), set ANTON_MINDS_URL in your environment or in ~/.anton/.env before running setup — setup honors it.

Option 2 — Bring your own key

Choosing "Bring your own key" presents four sub-options. In each case Anton validates the key with a quick probe call before saving it.

Anthropic

Enter your Anthropic API key and a model (default: claude-sonnet-4-6). Web search and web fetch run as Anthropic native server tools, billed on your Anthropic key — no extra setup.

OpenAI

Enter your OpenAI API key and a model (default: gpt-5.4). Web search runs natively through the OpenAI Responses API; fetching is covered by the same capability.

Google Gemini

Anton talks to Gemini through Google's OpenAI-compatible endpoint. Get a key at aistudio.google.com/apikey; the default model is gemini-3-flash-preview. Because this is a generic OpenAI-compatible endpoint, web search needs an external provider — run anton setup-search to configure Exa.ai or Brave Search.

Custom OpenAI-compatible endpoint

Works with Ollama, vLLM, Together, Groq, LM Studio, Azure, or any OpenAI-compatible API. You provide:

• Base URL — e.g. http://localhost:11434/v1 for Ollama
• API key — press Enter to skip if your endpoint doesn't need one
• Model name
• API version — leave blank for standard endpoints; required for Azure

After setup, Anton offers to configure a web search provider (Exa.ai or Brave Search) since generic endpoints don't expose search natively. You can run that step any time with anton setup-search — see Search providers.

Web search and fetch by provider

Anton exposes two web tools to the agent — web_search and web_fetch — both on by default. How they execute depends on your provider:

Provider	web_search	web_fetch	Setup
Anthropic BYOK	Anthropic native server tool	Anthropic native server tool	None — billed on your Anthropic key
OpenAI BYOK	OpenAI Responses API native	covered by web_search	None — billed on your OpenAI key
MindsHub	MindsHub passthrough	MindsHub passthrough	None — billed on your MindsHub key
Generic OpenAI-compatible (Together, Groq, Ollama, vLLM, …)	Exa.ai or Brave (you choose at setup)	stdlib HTTP GET (no key)	Run anton setup-search once

To opt out, set ANTON_WEB_SEARCH_ENABLED=false and/or ANTON_WEB_FETCH_ENABLED=false in your config. See Web search and Web fetch.

Switching providers later

• In chat: type /llm to change provider, model, or API key.
• From the terminal: run anton setup to re-run the full provider selection.

Both flows validate the new configuration before saving it to ~/.anton/.env.