Skip to content

Providers

SuperQode supports multiple provider types for connecting to AI models and agents.

Provider Types

  • BYOK Providers

    Bring Your Own Key - connect to cloud AI providers using your API keys.

    BYOK Providers

  • ACP Agents

    Agent Client Protocol - connect to coding agents like OpenCode, Amp, Claude Code, and more.

    ACP Agents

  • Local Providers

    Run models locally with DS4, Ollama, LM Studio, MLX, or vLLM.

    Local Providers

  • OpenResponses

    Use the OpenResponses gateway for enhanced capabilities.

    OpenResponses

  • Model Profiles

    Per-provider and per-model tuning: prompt suffixes, tool exclusions, init kwargs, pre-init hooks.

    Model Profiles

Execution Modes

SuperQode supports three execution modes:

Mode Description Use Case
BYOK Cloud AI providers with your API key Production, team use
ACP Coding agents with full capabilities Complex tasks
Local Self-hosted models Privacy, cost savings

Quick Start

BYOK (Cloud Providers)

# Set API key
export ANTHROPIC_API_KEY=sk-ant-...

# Connect
superqode connect byok anthropic claude-sonnet-4

ACP (Coding Agents)

# Connect to OpenCode
superqode connect acp opencode

# Connect to Amp
superqode connect acp amp

Local (Self-Hosted)

# Start a local OpenAI-compatible server.
# Example: Ollama
ollama serve

# Connect
superqode connect local ollama qwen3:8b

Provider Tiers

Tier 1 (First-Class Support)

Full support with optimized prompts (Enterprise adds prompt packs):

Provider Models
Anthropic Claude Opus 4.5, Sonnet 4, Haiku 4
OpenAI GPT-4o, o1
Google Gemini 2.5 Pro/Flash
Deepseek Deepseek V3, R1
Mistral Mistral Large
xAI Grok

Tier 2 (Supported)

Tested and supported:

Provider Models
(Model hosts and smaller providers) Various models

Local Tier

Self-hosted options:

Provider Description
DS4 Purpose-built DeepSeek V4 Flash server for local coding and long-context work
Ollama Easy local deployment
LM Studio GUI-based local models
MLX General Apple Silicon model serving
vLLM High-performance inference
SGLang Efficient structured generation
TGI Text Generation Inference
llama.cpp Lightweight CPU-first inference

Listing Providers

# List all providers
superqode providers list

# List by category
superqode providers list --category us
superqode providers list --category local

# Show configured only
superqode providers list --configured

Testing Connections

# Test a provider
superqode providers test anthropic

# Test with specific model
superqode providers test anthropic -m claude-sonnet-4

Configuration

In superqode.yaml

default:
  mode: byok
  provider: anthropic
  model: claude-sonnet-4

providers:
  anthropic:
    api_key_env: ANTHROPIC_API_KEY
    recommended_models:
      - claude-opus-4-5
      - claude-sonnet-4

  ollama:
    base_url: http://localhost:11434
    recommended_models:
      - qwen3:8b

Provider Categories

US Labs

Provider API Key Variable Documentation
anthropic ANTHROPIC_API_KEY docs.anthropic.com
openai OPENAI_API_KEY platform.openai.com
google GOOGLE_API_KEY ai.google.dev
xai XAI_API_KEY x.ai/api

China Labs

Provider API Key Variable
deepseek DEEPSEEK_API_KEY
zhipu ZHIPU_API_KEY
alibaba ALIBABA_API_KEY

Model Hosts

Provider API Key Variable Notes
openrouter OPENROUTER_API_KEY 95+ models
together TOGETHER_API_KEY Open models
groq GROQ_API_KEY Fast inference
fireworks FIREWORKS_API_KEY Open models

Local Providers

Provider Default Port Notes
ollama 11434 Easy setup
lmstudio 1234 GUI interface
mlx 8080 Apple Silicon
vllm 8000 Production
sglang 30000 Structured generation
tgi 8080 Hugging Face TGI
llamacpp 8080 Lightweight CPU inference
ds4 8000 Local DS4 server

Free Tiers

Some providers offer free access:

# List free providers
superqode providers list --category free
Provider Free Models
Google AI gemini-flash-latest (limited)
Groq llama-3.3-70b (rate limited)
OpenRouter Some open models

Next Steps