# LLM Providers

Connect Everruns to the model vendors you use. Configure a provider once, then manage which models your agents can run on.

[← Back to home](/)

## How provider management works

Providers are a settings surface in the operator console, not a hard-coded list. You configure a provider first — its credentials and endpoint — and then the models it exposes become available to agents. The settings layer covers providers, API keys, and members together, so model access is governed in the same place as the rest of the platform.

Agents reference models, not vendor SDKs. An agent runs on a harness default model unless it sets an optional override, which means you can change providers or models without rewriting agents.

## Supported providers

Everruns connects to major model vendors through a consistent provider interface:

- ![](/icons/providers/anthropic.svg) **[Anthropic](https://docs.everruns.com/providers/anthropic/)** — Claude models with extended thinking.
- ![](/icons/providers/openai.svg) **[OpenAI](https://docs.everruns.com/providers/openai/)** — GPT and o-series models via the Responses API.
- ![](/icons/providers/azure.svg) **[Azure OpenAI](https://docs.everruns.com/providers/azure-openai/)** — OpenAI models deployed in your Azure resource.
- ![](/icons/providers/bedrock.svg) **[AWS Bedrock](https://docs.everruns.com/providers/bedrock/)** — models hosted on Amazon Bedrock.
- ![](/icons/providers/gemini.svg) **[Google Gemini](https://docs.everruns.com/providers/gemini/)** — Gemini models with context caching.
- ![](/icons/providers/fireworks.svg) **[Fireworks AI](https://docs.everruns.com/providers/fireworks/)** — fast, low-cost inference for open models.
- ![](/icons/providers/microsoft.svg) **[Microsoft MAI](https://docs.everruns.com/providers/mai/)** — Microsoft MAI models via Azure AI Foundry.
- ![](/icons/providers/openrouter.svg) **[OpenRouter](https://docs.everruns.com/providers/openrouter/)** — one key for a large multi-vendor model catalog.

Because the provider layer is vendor-neutral, adding a new model is a configuration change rather than an integration project.

## API keys

Programmatic access is managed alongside providers in the operator console rather than hidden in separate tooling. The same settings layer issues API keys, manages members, and configures connections, so operators control models and access from one surface.

## Related

- [Capabilities](/platform/capabilities) — the tools and behaviors agents use on top of a model
- [Integrations](/platform/integrations) — external services Everruns connects to
- [Providers docs](https://docs.everruns.com/providers/) — configuring each provider, its credentials, and models