Send every prompt to the right model — automatically
Connect several LLMs — affordable and premium — and our API routes every prompt by task type and complexity. Power your apps and AI agents while cutting LLM spend 50–90% with no loss in quality.
Get your $1 Starter key
Enter your promo code and email. We mint a key and send it to you — and show it here once.
Why route at all?
One endpoint decides what each prompt needs. You keep full control of which models you run.
Cut cost 50–90%
Most prompts are easy. Stop paying flagship prices for “hello” and simple Q&A — route them to a cheap or local model and reserve the expensive one for genuinely hard work.
Instant & private
Routing is instant and adds no noticeable latency — even for agents firing hundreds of calls. Your prompt text is never stored, and nothing leaves your stack unless you opt in.
Zero lock-in
We return the task type and a complexity tier — you map them to whatever models you already run (OpenAI, Claude, llama.cpp, vLLM, Ollama…). Swap models any time.
Patent-pending accuracy
Our router doesn't just guess the task — it tells apart subtle, neighboring categories like business vs finance with 94% accuracy, so every prompt reaches the right model.
accuracy separating close task types — e.g. business vs finance
How it works
1 · Send a prompt
POST the text (or your own embedding) to /v1/route with your key.
2 · Get a decision
Receive the task type, calibrated confidence and a low / medium / high complexity tier — in one tiny JSON.
3 · Pick your model
Map the tier to your models: easy → cheap or local, hard → flagship. Done.
# one routing decision { "decision": { "task_label": "math", "confidence": 0.93, "complexity": { "tier": "low", "score": 0.27 } }, "scores": [ { "task_label": "math", "probability": 0.93 }, … ] }
Built for AI agents & orchestrators
Autonomous agents fire hundreds of LLM calls per task — most of them simple. Route each call to the right-sized model and cut agent cost 50–90% with no loss in quality.
Many models, one endpoint
Plug in several LLMs behind a single routing decision — cheap/local for easy steps, flagship for hard reasoning. No model-selection logic in your agent.
Tuned per task & complexity
Every call is classified by task type and complexity, so you control exactly which model handles what — and change it any time without touching agent code.
Drops into your framework
crewAI, AutoGPT, LangGraph, MCP orchestrators (Hermes) — anything that calls an LLM. One endpoint, full control of your model pool.
Simple, honest pricing
Start for $1 and scale as you grow. Every plan pays for itself many times over — the savings on your model bills dwarf the price.
- 100k included decisions
- 120/min rate limit
- $0.10 per 1k overage
- 1M included decisions
- 600/min rate limit
- $0.05 per 1k overage
- 10M included decisions
- 3000/min rate limit
- $0.02 per 1k overage
All plans include task-type + complexity routing. Optional oracle escalations (an advanced LLM resolves novel prompts) are billed at cost. Enterprise / on-prem: contact us.
Request a paid plan
Tell us a bit about your use case and we’ll get you set up.