One coding agent, any model
AgenC Core is an open source coding agent for your terminal. It has 25 model providers built in, from OpenAI and Anthropic to models running on your own machine. Pick one when you start, and switch whenever you want.
Start with any provider
Give AgenC your key for that provider and start it with --provider. The first-run wizard, agenc onboard, walks you through the same choice.
| Provider | Start it with | Key |
|---|---|---|
| xAI Grok | agenc --provider grok | XAI_API_KEY, or sign in with X |
| OpenAI | agenc --provider openai | OPENAI_API_KEY |
| Anthropic | agenc --provider anthropic | ANTHROPIC_API_KEY |
| Gemini | agenc --provider gemini | GEMINI_API_KEY |
| DeepSeek | agenc --provider deepseek | DEEPSEEK_API_KEY |
| Venice AI | agenc --provider venice | VENICE_API_KEY |
| OpenRouter | agenc --provider openrouter | OPENROUTER_API_KEY |
| Groq | agenc --provider groq | GROQ_API_KEY |
| Cerebras | agenc --provider cerebras | CEREBRAS_API_KEY |
| Mistral | agenc --provider mistral | MISTRAL_API_KEY |
| Kimi (Moonshot) | agenc --provider kimi | MOONSHOT_API_KEY |
| Z.AI | agenc --provider zai | ZAI_API_KEY |
| Z.AI Coding Plan | agenc --provider zai-coding-plan | ZAI_CODING_PLAN_API_KEY |
| QwenCloud | agenc --provider qwen | DASHSCOPE_API_KEY |
| QwenCloud Token Plan | agenc --provider qwen-token-plan | QWEN_TOKEN_PLAN_API_KEY |
| MiniMax | agenc --provider minimax | MINIMAX_API_KEY |
| Meta | agenc --provider meta | MODEL_API_KEY |
| NVIDIA NIM | agenc --provider nvidia-nim | NVIDIA_API_KEY |
| GitHub Copilot | agenc --provider github | GITHUB_TOKEN |
| Amazon Bedrock | agenc --provider amazon-bedrock | AWS access keys |
| Ollama Cloud | agenc --provider ollama-cloud | OLLAMA_API_KEY |
| Ollama, on your machine | agenc --provider ollama | none |
| LM Studio, on your machine | agenc --provider lmstudio | none |
| OpenAI-compatible server | agenc --provider openai-compatible | optional |
| AgenC, managed | agenc --provider agenc | sign in to AgenC |
Use a subscription you already pay for
- Grok: sign in with X and AgenC runs on your SuperGrok or X Premium subscription instead of a metered API key. How it works
- Z.AI Coding Plan and QwenCloud Token Plan have their own providers, so their plan keys work directly.
- GitHub Copilot: the
githubprovider uses the Copilot API with your GitHub token.
Run models on your own machine
Ollama, LM Studio and any OpenAI-compatible server, such as llama.cpp or vLLM, work with no API key. AgenC reads each model's context window from the server. Local models
Private models
With Venice AI you can pick models that Venice runs with zero data retention. AgenC lists Venice's live models and prices. AgenC with Venice
Switch whenever you want
Inside a session, /provider and /model change the provider and the model. When you start AgenC, --model also takes a provider and a model together:
agenc --provider anthropic agenc --model ollama:qwen3-coder:30b
To make a choice your default, put it in ~/.agenc/config.toml:
model_provider = "deepseek" model = "deepseek-flash"
DeepSeek V4.1 Flash is the setup behind AgenC's 76.2% on Terminal-Bench 2.1 for $34. AgenC with DeepSeek
Questions
Which AI models can AgenC Core use?
Any model from its 25 built-in providers, including OpenAI, Anthropic, xAI Grok, Gemini, DeepSeek, Mistral, Kimi, Qwen, Venice, OpenRouter, Groq and Cerebras, plus local models through Ollama, LM Studio or any OpenAI-compatible server.
Do I need an API key?
For a cloud provider you use your own key for that provider. Local models need no key, and Grok can run on your SuperGrok or X Premium subscription when you sign in with X.
Can I switch models in the middle of a project?
Yes. Use /provider and /model inside a session, or --provider and --model when you start AgenC.
Is AgenC Core free?
Yes. AgenC Core is open source under the MIT license. You pay only your model provider, or nothing with a local model.
Open source under the MIT license. Source code · All providers · Verify your download