Operations
Any model.
Your choice.
Use cloud AI services for production, local models for privacy, or any combination. Switch providers without changing a line of code. Automatic routing picks the right model for each job.
Any model.
COMPLEXITY-BASED ROUTING
Job Queue
SIMPLE linting fix
MEDIUM api endpoint
COMPLEX refactor DAG
SIMPLE doc update
MEDIUM test suite
→
Router
COMPLEXITY
ROUTER
score ≤ 0.3 → Haiku
score ≤ 0.7 → Sonnet
score > 0.7 → Opus
→
Model Lanes
HAIKU
fast · low cost · high throughput
< $0.01 / job
SONNET
balanced · best coding · medium cost
~$0.10 / job
OPUS
powerful · deepest reasoning · max cost
~$0.49 / job
5 PROVIDERS ANTHROPIC AWS BEDROCK OPENAI COMPAT OLLAMA CLAUDE CLI
① Jobs classified — each incoming job receives a complexity score (simple / medium / complex)
② Router directs traffic — Haiku for fast low-cost tasks, Sonnet for code, Opus for deep reasoning
③ 5 providers — Anthropic SDK, AWS Bedrock, OpenAI-Compatible (Groq, HuggingFace…), Ollama, Claude CLI
Intelligent model routing
Not every job needs the most expensive model. The platform automatically classifies job complexity and routes to the right model — fast models for simple work, powerful models for complex reasoning. The result: significant cost savings without sacrificing quality where it matters.
Zero vendor lock-in
Switch model providers with a single configuration change. No code modifications required.
Related Features
See model routing in action
Request access to see how automatic routing optimizes your AI spend.
Request Early Access