Every request has a best model.
One API for every major model and agent tool. chAIn sends each request to the cheapest option that clears your quality bar, fails over when a provider breaks, and puts every provider on one bill.
One endpoint, one key: api.chain.me/v1
Cheapest lane that clears 85 right now: —
0 tokens routed · simulated
Illustrative prices and quality scores that drift over time. They are not real provider figures.
One integration. Every model.
One endpoint and one key. Set the model to auto, set your quality bar, and chAIn handles routing, failover and billing.
from chain import Chain client = Chain(api_key=CHAIN_API_KEY) reply = client.chat( model="auto", # cheapest model that clears your bar quality_bar=85, # your floor, per request or per policy messages=messages,) # reply.model, reply.cost and reply.route tell you what chAIn choseIllustrative. SDKs, endpoint and parameter names are confirmed when access opens.
- 01
Integrate once
One endpoint and one key for every model and tool. Prompts, streaming and tool calls included.
- 02
Score in real time
Live price, latency, quality and uptime for every model and provider.
- 03
Route to the best price
Each call goes to the cheapest model that clears the quality bar you set.
- 04
Fail over, bill once
Errors and rate limits reroute automatically. Every provider lands on one invoice.
Your app → chAIn → every model and tool.
Agents need more than a language model. They speak, listen, search, run code and use a browser. chAIn routes those tool providers through the same key, bill and routing engine as the models, so one integration covers the whole agent.
Any product, agent or workflow
POST /v1/chat/completionsFoundation models
Think- OpenAI
- Anthropic
- Meta Llama
- Mistral
- DeepSeek
- xAI
- Open-weight hosts
Agent tools
Act- Voice
- Transcription
- Images
- Video
- Web search
- Doc parsing
- Code execution
- Browser automation
- Translation
Planned coverage at launch. New models are added the day they ship, and tools are billed on the same invoice as tokens.
Spend that routes itself.
Engineering keeps uptime. Finance keeps the savings. Security keeps control.
Savings compound as routing learns.
Routing learns which model wins for each workload, so the gap between your default and chAIn widens without a redeploy.
View as table
| Day | Default model only | chAIn auto | Saved |
|---|
Every request, priced before it's sent.
Each call is tagged with the model chAIn chose and what it cost. Spend rolls up by team, feature and model, and every provider lands on one invoice.
Each request carries its team, feature, chosen model and cost, so spend rolls up without extra instrumentation.
What engineering, finance and security each get.
Routing gets the savings. These controls make it safe to run in production.
| Capability | What it does | Owned by |
|---|---|---|
| Price-first routing | Every request goes to the cheapest model and provider that meets your bar. | Finance |
| Automatic failover | Outages, rate limits and model deprecations reroute before your users notice. | Engineering |
| Custom policies | Pin a model, set budgets, or route by region, latency or quality. The router only chooses among the options your policy allows. | Engineering |
| Cost analytics | Spend by team, feature and model, with savings tracked per request. | Finance |
| One key, one bill | Every provider behind one API key and one monthly invoice. | Finance |
| Enterprise controls | SSO, audit logs, spend limits, data residency and zero-data-retention routing. | Security |
Questions teams ask first.
Do I have to rewrite my code?
No. chAIn is one endpoint for every model and tool, with streaming and tool calls built in. Most apps switch by pointing their client at chAIn and choosing a routing policy.
Can I keep a specific model for some calls?
Yes. Policies let you pin a model, cap spend, or route by region, latency or quality. The router only chooses among the options your policy allows.
What happens when a provider goes down?
chAIn retries the request on the next-cheapest option that still meets your bar, so your app gets a response instead of an error.
Is chAIn tied to a model provider?
No. chAIn doesn't sell a model of its own, so it has nothing to steer you toward. Routing follows your price and quality settings.
How does chAIn make money?
A small routing fee on usage, annual plans for enterprise controls and, for the largest accounts, a share of the savings we deliver. Pricing is published when access opens.
When can I use it?
Join the waitlist and we'll email you when your access is ready.
The neutral router for the AI economy.
chAIn doesn't sell a model of its own, so routing follows your price and quality settings, not ours. Join the waitlist and we'll email you when access opens.
Prefer email? Write to info@chain.me.