Pricing

Published rates. No one is being billed yet.

These are the rates we will charge once public alpha opens. Hobby is and will remain free. PAYG and Team rates are indicative and may move before GA — we will publish a 60-day notice for any change.

Indicative rates · subject to change before public launch · no charges until alpha invites go out

Hobby

$0forever

Local agent. Your own compute. No charge.

✓ Tanvrit Compute SDK + CLI
✓ Local desktop agent (your laptop / homelab)
✓ Package and run jobs locally
✓ Job logs, status, retry — all local
✓ No cloud-tier usage
✓ Best-effort community support

Try locally

PAYG

$0.05/ CPU-hour

Managed compute on Tanvrit-operated infrastructure.

✓ Everything in Hobby
✓ Cloud-tier CPU pool ($0.05 / CPU-hour)
✓ GPU pool ($1.50 / GPU-hour, A10 equivalent)
✓ Per-second metering, billed monthly
✓ Email support, 1 business-day SLA
✓ 100 req/sec API rate limit

Available at public alpha

Reserve early access

Team

$199/ team / month + PAYG

Shared workspace, RBAC, audit logs.

✓ Everything in PAYG
✓ Up to 25 seats included
✓ Workspace-level RBAC (owner / admin / member)
✓ Audit log export (JSON, 90-day retention)
✓ Shared secrets vault
✓ 1000 req/sec API rate limit

Available at public alpha

Reserve early access

Enterprise

Customannual contract

On-prem, VPC peer, region pinning.

✓ Everything in Team
✓ On-prem agent fleet management
✓ VPC peering / private networking
✓ Region pinning (US, EU, AP)
✓ SSO via OIDC / SAML
✓ Custom rate limits, 99.9% SLA
✓ Named support engineer

Talk to sales

Pricing FAQ

Will the rates above change before public alpha?

They might — that is why we labelled them indicative. If we move a rate, we will publish a changelog entry, email everyone on the waitlist, and honour the published rate for 60 days for anyone who joined before the change.

Why is no one being billed yet?

Public alpha has not opened. The cloud-tier endpoints are stubbed and the billing meter is in dry-run. Hobby tier (local agent) is usable today and is free.

What counts as a CPU-hour?

One vCPU consumed for one hour, metered per second. Memory is included up to 4 GB per vCPU; beyond that, $0.005 / GB-hour.

What is an A10 equivalent?

One NVIDIA A10G or similar 24 GB GPU. We will publish a coefficient table for A100 / H100 / L40S / L4 before we open GPU access.

Can I prepay?

Not yet. We are not accepting payment until public alpha opens with a real Stripe meter against a real ledger.

All prices in USD. The local hobby tier never bills. Cloud-tier billing starts only after you explicitly accept a paid plan inside the portal — which is not live yet. Reserve early access →