Providers
200+ LLM providers. Switch without changing your workflow.
Use /provider inside the REPL to add, edit, or switch provider profiles at any time. No config files to edit manually — the wizard handles credentials, base URLs, and model selection.
/provider # add, edit, or switch provider profiles
/provider doctor # health check: reachability, auth, model availability
All providers
| Provider | Preset | Auth |
|---|---|---|
| Anthropic | anthropic | API key or OAuth |
| OpenAI | openai | API key |
| Google Gemini | gemini | API key |
| DeepSeek | deepseek | API key |
| Mistral | mistral | API key |
| xAI / Grok | xai-oauth | xAI OAuth |
| GitHub Copilot | github-onboard | GitHub OAuth |
| Codex (ChatGPT) | codex-oauth | ChatGPT OAuth |
| Groq | groq | API key |
| OpenRouter | openrouter | API key |
| Together AI | together | API key |
| Cloudflare Workers AI | cloudflare-workers-ai | API key |
| Cloudflare AI Gateway | cloudflare-ai-gateway | API key |
| NVIDIA NIM | nvidia-nim | API key |
| MiniMax | minimax | API key |
| Z.AI (GLM Coding Plan) | zai | API key |
| Moonshot AI | moonshotai | API key or Kimi Code OAuth |
| Azure OpenAI | azure-openai | API key |
| Alibaba Coding Plan | dashscope-intl | API key |
| Alibaba Coding Plan (China) | dashscope-cn | API key |
| Bankr | bankr | API key |
| OpenCode Zen | opencode-zen | API key |
| OpenCode GO | opencode-go | API key |
| AWS Bedrock | bedrock | AWS credential chain |
| Google Vertex AI | vertex | Application Default Credentials |
| Azure AI Foundry | foundry | DefaultAzureCredential |
| Ollama | ollama | None — local, free |
| LM Studio | lmstudio | None — local, free |
| Atomic Chat | atomic-chat | None — local, free |
| Any OpenAI-compatible | custom | Base URL + optional key |
Local models (Ollama)
Run Claudin fully offline with no API key using Ollama.
# Pull a model
ollama pull qwen2.5-coder:7b
# Launch claudin with it
ollama run qwen2.5-coder:7b
Any model available on ollama.com/library works — llama3, mistral, phi4, deepseek-coder, and more.
Advanced usage
Multi-profile workflows
/provider supports any number of profiles, switchable at any time — the change takes effect on the next request. Common patterns:
- Cost split — a strong cloud profile (Anthropic, OpenAI, Codex) for hard sessions and a fast/cheap profile (DeepSeek, Ollama, Groq) for routine tasks.
- Provider redundancy — primary cloud profile plus a fallback on a different provider for outage windows.
- Local vs remote — local Ollama profile for offline work, cloud profile for connectivity-dependent tasks.
- Role split — separate profiles for
devandreview, switched manually between sessions.
Profiles live under providerProfiles[] in ~/.claudin/settings.json; the active one is referenced by activeProviderProfileId. All sessions and sub-agents share the same active profile — there is no per-agent routing.
Custom headers
The Anthropic preset includes a Custom headers field on the profile (one Header: Value per line), sent on every request from that profile. Use it to route through proxies that require extra headers, or to tag requests at a corporate gateway.
/provider doctor
/provider doctor runs reachability, auth, and model-availability checks against the active profile — useful when a fresh profile fails to send its first message, a previously working profile starts erroring out, or you want a one-shot before/after comparison after editing a profile. It never mutates state, only reports.
/provider migrate vs. /import
/provider migrate and /import solve different problems. /provider migrate copies only your Claude Code sign-in — OAuth tokens and provider settings — and is offered automatically as a banner the first time you open /provider with an existing ~/.claude/. /import is the general form: it reads eight agents (Claude Code, openclaude, Codex, Gemini CLI, Qwen Code, opencode, Kimi CLI, Cursor), never touches credentials, and never overwrites anything without confirmation. See the migration guide for the full walkthrough.
Custom proxies & LiteLLM
The custom preset works with any OpenAI-compatible base URL — self-hosted gateways, corporate proxies, or LiteLLM. Set baseUrl, optionally apiKey, and the model alias your gateway exposes.