You pay the provider price for the models you choose, plus $0.25 per million tokens your agents process. Calls, services, and retained data appear only when you use them. Everything else is included.
Plans change support and commitments — never the rate card and never how work executes. Every plan runs the complete Agent OS.
The full product at published rates. Continuity is a card, not a plan change.
Same product, same rates. The fee buys published commitments — and credits if we miss them.
Negotiated economics and operational guarantees over the same rate card.
The latest models through one API, with routing, retries, failover, and caching built in.
Model usage is charged at the published rate for each model. Noorle adds no surcharge to input, output, or cached tokens. We currently support the latest models from OpenAI and Anthropic, alongside leading open-weight models.
One platform rate on every token the Agent OS processes — the same across every provider and model, wherever the processing happens.
A call is one recorded instruction to the platform: a run starting, or a tool being dispatched. Every call costs the same. What the call goes on to do bills on its own line.
Runtimes and outside-world capabilities, each with one published unit and one price. A resource is billed on one line — never stacked meters.
| Compute | Shape | Price |
|---|---|---|
| Code Runner | Ephemeral sandboxed execution · whole seconds · one-second floor | $0.00003 / second |
| Plugin (WASM) | Ephemeral wasm isolate | $0.00003 / second |
| Sandbox | Full container · isolated execution environment | $0.000045 / second |
| Browser | Interactive session · metered per millisecond · no floor | $0.10 / hour |
| Computer | X2 → X32 · 2–32 GB RAM · 2–16 vCPU · price scales linearly with size | $0.06 – $0.96 / hour |
| Service | Billing unit | Price |
|---|---|---|
| Web search · standard | 1,000 queries | $20.00 |
| Web search · deep | 1,000 queries | $40.00 |
| Knowledge search | query | $0.00002 |
| Knowledge indexing | 1M tokens processed | $0.20 |
| SMS | message | $0.015 |
| Phone number | month | $3.00 |
| Channels · Slack, Telegram, email | message | $0 |
| HTTP client · Files · Connector | request | $0 |
Charged only for data you intentionally keep beyond the plan’s included window. Deleted by default at the end of it.
| Class | Billing unit | Price |
|---|---|---|
| Standard | GB / month · files, run history, logs, archives | $0.15 |
| Indexed | GB / month · vector and search indexes, knowledge bases | $1.50 |
Included covers objects, routing, waiting, and recovery — never the instruction. We charge for intelligence, calls, and resources; not architecture, seats, or idle time.
Cost control belongs at the execution boundary, close to the work—not in a spreadsheet at month end.
See a cost range before work starts. An enforceable maximum beats a falsely exact estimate.
Cap spend by organization, client, agent, workflow, or a single run. Work checkpoints cleanly at the limit.
An append-only ledger attributes every charge, with provider and Noorle lines kept separate.
One dollar balance funds everything. Auto-reload with thresholds, maximums, and notifications.
Add a card — we won’t charge you — and start with $10. No seat fees. No subscription required to start.
