Your providers. Your choice.

Start with a verified preset

Max Harness includes 25 provider presets researched against official documentation. A preset sets the API address and adapter. You supply your own credentials and remain responsible for provider charges and account access.

Use Fetch models to retrieve a model list when the provider supports discovery. Some providers publish a public list, so successful discovery does not necessarily prove your key works. Test connection sends a small generation request and may use credits.

Model support varies

Image input, PDF input, tools, generation settings and usage reporting depend on the model and its API route. Capability labels are conservative. A manually entered model starts without advanced capabilities until you explicitly confirm them from its documentation.

Some services route models through different API formats. Max Harness keeps that route with the discovered model where the provider supplies it. A coding subscription does not automatically include general API access.

Add a custom endpoint

Use a custom OpenAI-compatible, OpenAI Responses, Anthropic or Gemini endpoint when its API follows the implemented format. The server blocks private network destinations unless an administrator has allowed an exact host and port. This makes a local Ollama connection an intentional server configuration choice.

Read the connection guide for credentials, errors and model setup.