Skip to content

Models and inference

These routes assign ownership of tools, context, memory, approvals, evaluation, and workflow to the SuperQode harness.

Hosted model gateways

Integration Dependency and authentication Start Verify Detailed guide
LiteLLM gateway Included; set the selected provider's API key :connect byok <provider> <model> superqode providers doctor <provider> BYOK Providers
OpenResponses gateway Included; configure a compatible endpoint and credential Select openresponses in superqode.yaml superqode config validate OpenResponses
Hugging Face Hub uv tool install "superqode[hf]"; use HF_TOKEN for gated models superqode models hub <query> superqode models show <model> Model Catalog
Exa search client uv tool install "superqode[extras]", then set the Exa credential Enable an Exa-backed search route Run the configured search tool Airplane Mode and External Search

The provider catalog includes OpenAI, Anthropic, Google, xAI, DeepSeek, Mistral, Moonshot, Qwen, Z.AI, MiniMax, OpenRouter, Together, Groq, Fireworks, Hugging Face, Cerebras, Cohere, Bedrock, Azure, Vertex AI, Cloudflare, and other supported LiteLLM or OpenAI-compatible routes. Query the live catalog instead of copying a static model identifier:

superqode providers list
superqode providers models <provider>
superqode providers doctor <provider>

See Providers for provider tiers, authentication, and the complete model-provider catalog.

Local inference

Integration Install or service Start Verify Detailed guide
Ollama Install and start Ollama :connect local ollama <model> superqode local doctor Local Providers
LM Studio Start the local OpenAI-compatible server :connect local lmstudio <model> superqode local doctor Local Providers
vLLM Start a vLLM OpenAI-compatible endpoint :connect local vllm <model> superqode local doctor Local Providers
SGLang Start an SGLang endpoint :connect local sglang <model> superqode local doctor Local Providers
Hugging Face TGI Start a TGI endpoint :connect local tgi <model> superqode local doctor Local Providers
MLX and MLX-VLM uv tool install "superqode[mlx]" on supported Apple Silicon :connect local mlx <model> superqode local doctor Local Providers
DwarfStar Install and start a compatible DwarfStar server Select the discovered local route superqode local doctor Local Providers
Apple Foundation Models Install apple-fm-sdk on supported Apple Silicon with Apple Intelligence Set SUPERQODE_UTILITY_PROVIDER=apple-fm superqode local doctor Local Stack Doctor
Transformers Install transformers, accelerate, and the required compute framework Select the local Transformers route superqode local doctor Local Stack Doctor

Generate a repository-owned local harness after the server is ready:

superqode local init --repo .
superqode --harness superqode.local.yaml