The AI router for greener, sovereign AI.
One API to the major LLM providers — hosted in Europe, sovereignty on your terms, with a carbon receipt on every call. We're a proxy: your prompts pass through to the provider you chose. Never stored, never transformed, never trained on.
Paste it into the custom-provider screen of Cursor, Claude Code, OpenCode, Aider, Cline, or Goose. No plugin, no proprietary protocol.
One hop. Fully accounted for.
Your app or agent
LowRouter
Provider you chose
auto mode: eligible models → lowest carbon → lowest cost → round-robin
Every response returns with lowrouter_metadata: provider, region, gCO₂eq emitted, and grid carbon intensity. Cost per call lands in your dashboard. The route is proven, not promised.
Many AI providers. One invoice.
Stop juggling provider accounts, prepaid balances, and stranded credits. You pay LowRouter; we pay the providers.
- One balance, pay-as-you-go — top up when you need, no seats, no commitment.
- Token usage billed at the upstream provider's price, no per-call markup. The only fee is the processing fee at top-up, shown before you confirm.
- One EU VAT-compliant invoice for finance, whatever mix of providers you used.
What LowRouter records for every request
Your dashboard breaks down usage across every provider you route to: tokens in and out, upstream cost, and gCO₂eq emitted. Export it as CSV — auditable, per-request data your ESG team can put toward Scope 3 reporting.
Finance reporting and sustainability reporting, from the same source of truth.
| Model | Provider | Tokens | Cost | gCO₂eq |
|---|---|---|---|---|
| mistral-large | mistral | 1,204,880 | €3.10 | 41.2 |
| mistral-small | scaleway | 2,051,300 | €1.28 | 22.7 |
| claude-haiku-4.5 | anthropic | 838,540 | €1.90 | 78.5 |
| gpt-4o-mini | openai | 521,010 | €0.74 | 66.0 |
| llama-3.3-70b | groq | 149,200 | €0.11 | 9.4 |
| Total | 4,764,930 | €7.13 | 217.8 |
Keep your existing OpenAI client
There is no LowRouter SDK and no proprietary protocol. Point anything OpenAI-compatible at https://api.lowrouter.ai/v1.
- SDKs
- OpenAI PythonOpenAI NodeLangChainLiteLLMany OpenAI-compatible…
- Coding agents
- Claude CodeOpenCodeGooseAiderClineCursor…
Switch back any time. It's just a base URL.
Aliases
Call alias/my-model in your code; repoint it to any provider, model, or region later — zero client changes.
Auto-routing
Coming soonSet a policy once — e.g. a carbon budget per key — and let LowRouter pick the route that fits, per request.
A sovereign platform, with carbon estimates you can audit
🇪🇺 Sovereign
The platform: LowRouter runs in France, operated by Carbonifer SAS (Nantes). GDPR-native, outside CLOUD Act reach. Your keys, billing data, and usage metadata never touch US jurisdiction.
The inference: your choice. Filter the catalogue to EU-only providers and pin an alias to a specific provider, model, and region. The region we routed to is reported in every response.
🌱 Carbon transparency
Per-request gCO₂eq in every response — computed from open public grid data (Ember Climate, refined by regional grid operators: RTE, PJM, CAISO, National Grid ESO, AEMO…) and the EcoLogits energy model.
Pick the lowest-carbon route. Compare providers by carbon intensity per token in the catalogue and choose the cleanest one that fits — or let lowrouter/auto do it for you.
These are estimates. They come from regional grid averages applied to modeled inference energy — not lifecycle emissions, and not meter readings from the provider's datacenter. When we can't classify a request, we report unknown rather than fabricate a number. Read the methodology and what it deliberately leaves out.
From your first request to an auditable receipt
- 01
Point your SDK at us
Change the base URL to api.lowrouter.ai/v1. Done.
- 02
Pick your route
A provider-prefixed model (mistral/mistral/mistral-large), lowrouter/auto for the lowest-carbon route available, or one of your aliases. Compare carbon, cost, and jurisdiction in the catalogue first.
- 03
Read the receipt
Every response carries lowrouter_metadata: provider, region, gCO₂eq, grid intensity, cost. Log it, audit it, show it to your users.
Usage-based pricing, without seats or commitments
Token usage is billed at the upstream provider's price, with no per-call markup. The only fee is a processing fee applied when you top up credits, covering payment processing and our margin. No monthly commitment, no per-seat charge.
Top up credits when you need them, and the exact processing fee is shown before you confirm. Your dashboard then breaks down tokens, cost, and gCO₂eq per model and provider you've used, so finance and sustainability reporting come from the same source of truth.
What we claim, and where it's proven
| Claim | Where it's proven | Docs |
|---|---|---|
| Carbon transparency | Every response carries lowrouter_metadata with the region, grid carbon intensity, and gCO₂eq of that call — and the formula behind it is published. | Methodology → |
| EU sovereign | The platform runs in France, operated by Carbonifer SAS (Nantes). Your keys, billing data, and usage metadata stay in EU jurisdiction; the routed region is returned on every response. | Sovereignty → |
| One gate, one bill | Many providers behind one OpenAI-compatible API. One balance, one EU VAT invoice — tokens billed at the upstream price, the only fee at top-up. | Billing → |
| No lock-in | It's just a base URL. Call an alias and repoint it to any provider, model, or region later — or point your SDK back at a provider direct, any time. | Integration → |
Why we built this
We kept having to answer "what does our AI actually emit?" with vendor marketing instead of data. Carbonifer's job is measuring software's footprint, and the router is the one hop every request already crosses — so that's where we put the meter. If a number on this page can't be traced to a source, we'd rather not show it.
— Olivier, founder
Design decisions
- Missing carbon data is reported as unknown, never as an average. An estimate you can't trace is worse than a gap you can see.
- Sovereignty constraints beat carbon optimization. If you pin EU-only, we route EU-only — even when a lower-carbon region exists elsewhere.
- Repointing an alias trades reproducibility for flexibility. That's your call to make, so it's a named, visible action — never something we do silently.
Frequently asked questions
More questions? See the full FAQ in the docs.
What is LowRouter?
An AI router (a gateway/proxy): one OpenAI-compatible API in front of many LLM providers. We forward your request to the provider you chose and return the response with a carbon and cost receipt. We don't run models ourselves.
Are you an LLM provider?
No. We're a proxy. Your prompt goes to the provider you picked, streaming intact. We don't read, store, transform, or train on it.
Why do you list US and non-green providers at all?
Because you can't choose responsibly without seeing the alternatives. LowRouter shows the region, jurisdiction, and carbon cost of every route — including the non-sovereign and carbon-heavy ones — so the comparison is honest and the choice is yours. Sometimes you need a specific US model; use it with eyes open, and see exactly what it emits. And by putting every provider on the same transparent scoreboard, we give them a reason to compete on it: the day a provider ships greener, more sovereign inference, it shows up here — measured, not marketed.
Can I pin requests to a specific country or jurisdiction?
Yes — pin an alias to a provider, model, and region, or filter the catalogue to EU-only. The routed region is returned in every response.
How accurate are the carbon numbers?
They're estimates built exclusively on open, citable sources: Ember Climate country averages refined with regional grid-operator data, and EcoLogits for inference energy. Every assumption is documented in the methodology. We'd rather give you a defensible estimate than a marketing number.
Do you use my data?
No. No training, no analytics on prompt content, no resale. Requests are forwarded and forgotten; only billing metadata (tokens, model, cost, carbon) is kept for your dashboard.
How does pricing work?
Upstream provider price on tokens, zero per-call markup, one processing fee at credit top-up (shown before you pay).