Models & Providers
Choose between AgentOne's own model, your own API keys for 125+ providers, local models, and custom OpenAI-compatible APIs.
AgentOne works with almost any AI model. There are four ways to connect one, and you can use several at once:
| Option | You need | Cost |
|---|---|---|
| AgentOne's model | An AgentOne account | Free monthly credits, more with a paid plan |
| Built-in providers | An API key from the provider | Billed by the provider, no markup from AgentOne |
| Local models | Ollama or LM Studio | Free, runs on your computer |
| Custom providers | Any OpenAI-compatible API | Depends on the API |
AgentOne's Model
When you sign in, AgentOne's own model appears in the model selector as Default under AgentOne. It works right away, with no API keys, and routes each request to a capable model behind the scenes. It uses your account's monthly credits.
If you'd rather only use your own providers, turn on Settings -> Account -> Hide AgentOne models.
Bring Your Own Keys
AgentOne has more than 125 providers built in, including OpenAI, Anthropic, Google, xAI, Mistral, Groq, DeepSeek, OpenRouter, Together AI, Fireworks AI, Cerebras, Hugging Face, Vercel AI Gateway, and many more. Turn one on, paste your API key, and its models appear in the model selector. Your messages go straight from your computer to the provider.
See Built-in Providers.
Where Model Information Comes From
AgentOne knows each model's name, context window, pricing, and features (like tool and image support) from the AI Model Directory, an open, daily-updated catalog of over 22,000 model listings. AgentOne downloads the list on its own and keeps it fresh. See Model Metadata.