- Conduit
- Provider guides
Provider guides
Your API key, in a real desktop app
Conduit is a free, open-source desktop AI client that talks to seventeen model providers with your own key, or to a model running on your own machine with no key at all. These guides cover one provider each: how to set it up, what it costs, which Conduit features work with it, and where it falls short.
Guides
- OpenRouter — one key for hundreds of models from many companies, including models no chat subscription offers. Setup, the 5.5% credit fee, free-model limits, and web search billed to the same key. Checked 4 October 2026.
- DeepSeek — low-cost models that think before they answer. Setup, peak and off-peak prices, tools in thinking mode, where DeepSeek stores data, and what Conduit doesn't do with DeepSeek yet. Checked 4 October 2026.
More guides are on the way. Every provider below already works in Conduit, and the setup is the same for all of them.
Set up any provider in four steps
- Create an API key on the provider's website. Local providers (Ollama, LM Studio) need none.
- In Conduit, open Settings → Providers & keys and pick the provider from the Provider list.
- Paste the key into Provider secret and choose Save provider key. Conduit stores it in your operating system's keychain; the interface never sees it again.
- Choose Load models and pick one from the Model list. Conduit lists whatever models the provider offers your key.
You can keep keys for several providers and switch between them mid-conversation with Ctrl+Shift+P (⌘+Shift+P on macOS), or pick any model from the model menu under the message box.
Every provider Conduit supports
| Provider | What you need | Custom endpoint |
|---|---|---|
| OpenRouter | API key | No |
| Anthropic | API key | Yes |
| OpenAI | API key | Yes |
| Google Gemini | API key | No |
| xAI | API key | Yes |
| DeepSeek | API key | No |
| Z.ai | API key | Yes |
| Moonshot AI | API key | Yes |
| Qwen | API key | Yes |
| Mistral | API key | No |
| Groq | API key | No |
| Together AI | API key | Yes |
| Fireworks AI | API key | Yes |
| OpenCode Zen | API key | No |
| Ollama | Ollama running on your machine; no key | Yes |
| LM Studio | LM Studio's server running; key optional | Yes |
| OpenAI Compatible | Any server that speaks the OpenAI API, such as vLLM or LiteLLM; key optional | Yes |
"Custom endpoint" means you can change the address Conduit sends requests to, for instance to route Anthropic or OpenAI traffic through a LiteLLM proxy. MiniMax has no provider of its own in Conduit; reach it through OpenRouter, OpenCode Zen, or an OpenAI-compatible endpoint.
What Conduit adds, and what it doesn't
Conduit is free and open source (AGPL-3.0), with no account and no subscription. Requests go from your machine straight to the provider you picked; Conduit does not proxy them, resell tokens, or add anything to what the provider charges. What you get on top of the raw API is a desktop app: MCP tools that ask before they act, answers cited from your own files, runs you can steer while they work, and conversations stored on your own disk. Whether paying per token beats a monthly plan depends on how much you use it — the API vs subscription calculator works it out from your own numbers.
Running models locally instead? See local models with a real app around them. Weighing Conduit against another client? See all comparisons. Everything we publish is on the resources page.
The provider list and the custom-endpoint column match Conduit v1.0.0-rc.7 and are checkable in the source. Found an error? Tell us and we will fix it.