LLM providers

Cascade abstracts the LLM behind a single interface, so swapping providers is one config command. Each provider has its own setup story.

Anthropic Claude

The default. Uses Anthropic’s tool-use feature for reliable structured output.

cascade configure llm anthropic --key sk-ant-xxx --set-default

Setting	Value
Default model	claude-opus-4-7
Auth	API key
Env var	ANTHROPIC_API_KEY
Best for	Maximum reliability of structured output

OpenAI

Uses OpenAI’s Structured Outputs (response_format with json_schema). Works with any OpenAI-compatible endpoint via --base-url.

cascade configure llm openai --key sk-xxx --set-default
# Azure OpenAI:
cascade configure llm openai --key sk-xxx --base-url https://your.openai.azure.com/v1

Setting	Value
Default model	gpt-5
Auth	API key
Env var	OPENAI_API_KEY
Best for	Teams already on OpenAI; Azure OpenAI deployments

Google Gemini

Uses response_schema with Pydantic models for structured output.

cascade configure llm google --key gemini-xxx --set-default

Setting	Value
Default model	gemini-2.0-flash
Auth	API key
Env vars	GOOGLE_API_KEY or GEMINI_API_KEY
Best for	Teams on Google Cloud; cost-sensitive use cases

Claude Code SDK

Uses your local Claude Code subscription. No separate API key.

pip install "cascade-agent[claude-code]"
cascade configure llm claude_code --set-default

Setting	Value
Default model	claude-opus-4-7
Auth	Your Claude Code installation
Env var	None
Best for	Zero-setup adoption; teams with Claude subscriptions

Full guide: run without an API key →

Ollama / vLLM (local)

Any OpenAI-compatible local LLM server.

cascade configure llm ollama --model llama3.1 --set-default
# Custom endpoint:
cascade configure llm ollama --base-url http://gpu-host:8000/v1 --set-default

Setting	Value
Default model	llama3.1
Default base URL	http://localhost:11434/v1
Auth	None
Env var	None
Best for	Air-gapped environments; data residency requirements

Switching providers

Per command:

cascade prompt "Add health endpoint" --model claude-opus-4-7

Per project (via cascade.yaml):

agent:
  provider: openai
  model: gpt-5

Globally (default):

cascade configure llm anthropic --set-default

Coming in v0.2

• GitHub Copilot CLI (best-effort; Copilot’s API isn’t structured-output-friendly)
• OpenRouter as a one-config gateway to all providers
• Per-stage routing: planner on Gemini, coder on Claude, etc.