RouteIQ sits between your app and your AI providers. Simple queries go to cheap models. Complex ones go to powerful models. Automatically.
Your app calls RouteIQ exactly like it calls Anthropic or OpenAI. We handle the routing invisibly, in under a millisecond.
Tell us about your current AI usage and we’ll show your projected monthly saving with RouteIQ.
Based on live RouteIQ routing thresholds
Works with Anthropic, OpenAI, Groq and more. Your existing API keys work. Your existing model calls work.
# Direct to Anthropic client = Anthropic( api_key="sk-ant-...", )
# Via RouteIQ client = Anthropic( api_key="sk-ant-...", base_url="https://routeiq-production.up.railway.app", )
Works with any provider
Built for engineering teams who want meaningful savings without adding complexity.
Scores each query for complexity in under 1ms. Gets smarter with every query over time.
See exactly how much you’ve saved, which models were used, and a per-query routing breakdown.
Full server-sent events support. Token streams proxied in real time — zero latency difference for your users.
UK/EU native from day one. Data Processing Agreement included for regulated industries.
Set your own routing rules. Tell RouteIQ to never route legal queries below Sonnet.
If a provider goes down, RouteIQ retries via another automatically. More reliable than direct API calls.
An honest view of where each tool sits.
| Feature | RouteIQ | Martian | Requesty |
|---|---|---|---|
| Intelligent complexity routing | ✓ | ✓ | ✗ Manual only |
| EU / UK data residency | ✓ Native | ✗ US only | ✓ |
| SMB / startup pricing | ✓ | ✗ Enterprise | ✓ |
| Provider-neutral (any LLM) | ✓ | ✓ | ✓ |
| GDPR DPA included | ✓ | ✗ | ✗ |
| One-line integration | ✓ | ✗ | ✓ |
No — RouteIQ works with any AI provider. It currently supports Anthropic (Haiku, Sonnet, Opus) and OpenAI (GPT-4o mini, GPT-4o, GPT-4 Turbo). Groq, Gemini and Mistral are on the roadmap.
Yes — the classifier needs to read the query to score its complexity. All queries are processed in the EU, covered under GDPR, and never used to train models without consent.
You can set minimum model thresholds per use case. The classifier also improves over time as it learns your specific query patterns.
The routing decision takes under 1ms. RouteIQ often reduces perceived latency by routing simple queries to faster, cheaper models like Haiku which respond 3–5x quicker than Opus.
RouteIQ includes automatic fallback routing. If Anthropic is unavailable we retry via OpenAI automatically — making RouteIQ more reliable than a direct API call.
Yes. You keep your existing keys and billing relationship with providers. RouteIQ acts as a neutral proxy and you only pay us a percentage of the saving we generate.
Join the early access list. We’re onboarding the first 20 customers now.
Get started free →