AI providers¶

mr-review supports two AI backends: Anthropic (Claude) and any OpenAI-compatible API.

Anthropic (Claude)¶

The default provider. Uses the Anthropic Messages API with streaming.

Recommended models:

Model	Notes
`claude-sonnet-4-6`	Best balance of quality and speed (default)
`claude-opus-4-7`	Highest quality, slower
`claude-haiku-4-5-20251001`	Fastest, good for short diffs

Configuration:

AI_PROVIDER=anthropic
AI_API_KEY=sk-ant-...
AI_MODEL=claude-sonnet-4-6

Get an API key at console.anthropic.com.

OpenAI-compatible¶

Any endpoint that implements the OpenAI chat completions API with streaming.

OpenAI:

AI_PROVIDER=openai
AI_API_KEY=sk-...
AI_MODEL=gpt-4o

Self-hosted (Ollama):

AI_PROVIDER=openai
AI_BASE_URL=http://localhost:11434/v1
AI_API_KEY=ollama
AI_MODEL=qwen2.5-coder:32b

Other compatible services (Groq, Together AI, etc.) — set AI_BASE_URL to their API endpoint.

Choosing a model¶

For code review, models with large context windows perform best. The diff, pinned lines, commit message, and review brief are all included in the prompt — large MRs can exceed 20k tokens.