Free tier · CarbonLayer
Validate inference trade-offs with simulation.
Explore model-based cost and latency scenarios in the separate Playground, or try a simulated POST /api/inference receipt.
Without provider credentials, the Playground returns simulated output. When credentials are configured, it can make one direct provider call for a single request. That call is not production workload routing. By contrast, POST /api/inference returns a simulated decision receipt with illustrative route-target and latency metadata; it makes no provider call and reports no measured cost or environmental results.
The current site listing states that Free is $0/month with a 50K monthly receipt allowance and no card; current availability and commercial terms are not independently verified in this update. Requests appear in the key-scoped status log at /api/savings. The log does not represent production tracking.
The route log does not report environmental savings. Read Methodology v1.0 for the evidence framework and current unavailable impact fields.
Already experimenting? Create a free account and explore the prototype dispatch log with its explicit unavailable-impact status.
What you get
Built around three things AI engineers ask for.
Per-request simulated receipts
Every POST to /api/inference persists request attribution, a simulated model label, a seeded route target, and latency. Impact data is unavailable in this prototype.
Impact evidence status
Environmental data is unavailable in the current prototype. A future badge concept does not contain measured carbon, water, or an inferred per-request impact value.
Scheduler planned
The route-window preview is illustrative simulation metadata. Scheduler execution and physical workload shifts are planned capabilities, not available behavior.
Founder’s Circle / Beta Access Plans
Start free, explore the prototype.
Availability-qualified previews. Existing displayed prices and monthly inference allowances are preserved, but current commercial terms are not independently verified in this update. The allowance counts describe calls to the simulated receipt API, not model execution or a delivery service. Scheduler execution, live Routing, and CSRD reporting are planned. Visit /pricing for the full breakdown.
Free tier
For prototypes and engineers exploring simulated receipt and route-target metadata.
- 50K simulated receipt API calls / month
- Per-call simulated route receipts
- Illustrative scheduling preview; scheduler execution planned
- Key-scoped receipt log read
50K simulated receipt API calls / month. Badge and priority-capacity features are not verified production entitlements.
Team preview
For teams exploring higher-volume prototype workloads. Production capacity and audit reporting are planned or configuration-dependent.
- 25M simulated receipt API calls / month included
- Simulated region-target metadata
- Simulated route-target preference; live Routing planned
- Planned audit export
Try the simulated API request
No SDK, no wrapper, no shim layer.
POST a request from curl and get back a stable simulated receipt with the route target, latency, and explicit provider/impact status.
curl -sX POST {{api}}/api/inference \
-H "authorization: Bearer cai_…" \
-H "content-type: application/json" \
-d '{
"prompt": "Suggest 3 ways to halve the warehouse's overnight energy bill",
"model": "claude-haiku",
"tokens": 600,
"preferLowLatency": false
}'Prototype: this endpoint does not call Claude or any other language-model provider. response echoes up to 80 characters of the trimmed prompt; model is a simulated label, and impact data is unavailable.
Not ready to sign up?
Stay close to the work, on your terms.
Choose the lightest-touch way to keep up: one email when something material changes, or a follow for the day-to-day build in public.
Stay in the loop
Get the material updates.
One message when the free tier widens, a new region opens, or dispatch policy changes. Nothing in between.
Follow the build
See CarbonLayer in motion.
Follow CarbonLayer on X for product progress, dispatch notes, and the ideas shaping more transparent AI infrastructure.
Follow @CarbonLayerHQThe listed 50K monthly inference allowance and paid plan amounts are preserved from the existing site copy, not independently verified in this update. Pro and Team previews are availability/configuration-dependent; Scheduler execution, live Routing, CSRD reporting, audit exports, and a generated embed badge are planned. Inference allowances do not describe evaluation or delivery services.