Find a plan to power your agentic solution.
Start free, add prepaid credits when you need more, and only talk to sales when you need to. Every tier runs on the same Firecracker microVMs, the same APIs, and the same 5M-per-project ceiling.
Usage Credits put you in control: start with a $5, $10, or $20 prepaid pack, see your live balance and activity, and turn on auto-refill only when you want it.
Start with creditsFree
For developers kicking the tires.
1 vCPU · 1 GB RAM · 10 GB disk
Per sandbox session
Self-serve docs, community Slack
Encrypted storage, 7 day retention
Usage Credits
Prepaid flexibility without surprise overages.
Up to 4 vCPU · 16 GB RAM · 100 GB disk
Per sandbox session
500–2,000 CUs · $0.01 per CU · no postpaid overage
Live balance, activity, and spend control
Pro
For production agents at scale.
Includes 25,000 CU ($250 of usage) per billing cycle. Additional usage is billed at $0.01 per CU at the Pro rates below. The self-serve Pro billing cycle is monthly.
Up to 16 vCPU · 64 GB RAM · 100 GB disk · snapshot + resume
Per sandbox session
40% lower rates · $0.01/CU overage
P1 SLA: 24 hours
Enterprise
Custom scale, VPC, and compliance.
Custom vCPU · RAM · disk ceilings
Per sandbox session
SSO / SAML · RBAC · DPA
P1 SLA: 1 hour · Resident SA
Credit packs.
Pick a pack when you join Usage Credits. Top up at any time, buy up to 10 packs at once, or set threshold-based auto-refill.
1 CU = $0.01 of prepaid balance. Credits pay for sandbox CPU, memory, disk, and snapshot storage at the rates below. Network egress is free.
Rates.
Usage Credits deduct from your prepaid balance at these per-resource rates; Pro draws the same resources from its included 25,000 CU first, then bills overage at $0.01/CU. CPU tracks active runtime execution when metering is available, rather than reserved vCPU time. RAM and disk are allocation-based while a sandbox is running, and snapshot storage is billed per GB per 30-day month while suspended.
Pricing calculator.
Describe a typical month of sandbox activity. We convert it to compute units with the production rate sets and show what it costs On-Demand and on Pro, including Pro's $250 of included usage.
1,250 sandbox-hours a month. On-Demand costs $117.50 lessat this volume. Pro's $250 base pays for itself at about 5,661 sessions a month of this shape.
Estimates before tax. Usage is pooled per organization and billed in whole compute units. Snapshot storage uses a fixed 30-day month. Active CPU falls back to allocated vCPU when active metering is unavailable. Rate sets: usage_credits_gb_v1, pro_gb_v1.
Frequently asked.
If you have a question that isn't covered here, email support@tensorlake.ai — a founder will reply.
Usage Credits are a prepaid balance for sandbox compute. One compute unit (CU) is worth $0.01. Choose a $5, $10, or $20 pack when you join the plan, then add more whenever you need them. CPU, memory, disk, and snapshot storage deduct from that balance at the rates shown above — never as a postpaid overage.
Pro is $250 per billing cycle and includes 25,000 CU, which is $250 of usage at Pro's lower per-resource rates. Usage beyond the included 25,000 CU is billed at $0.01 per CU on the same invoice. Pro also lifts limits to 1,000 concurrent sandboxes, 16 vCPU and 64 GB RAM per sandbox, and unlimited session duration.
You can start on our Free tier and run 1 sandbox at a time — no credit card required. When you're ready to scale past the Free limits, switch to Usage Credits and choose a prepaid pack starting at $5.
We block new sandbox activity and stop running sandboxes instead of turning the shortfall into an unexpected bill. You can buy another pack to resume access, or enable auto-refill with your preferred pack and balance threshold to reduce interruptions.
Yes. We are SOC 2 Type 2 compliant on every tier, and HIPAA compliant with a BAA on Pro and above. For strict data residency or sovereignty requirements we offer on-prem and in-VPC deployment on Enterprise — email support@tensorlake.ai to start a conversation.
CPU usage is billed from active runtime execution when active metering is available, rather than from reserved vCPU time. If an active CPU sample is unavailable or untrustworthy, we use allocated vCPU time for that sandbox generation. RAM and disk are allocation-based while the sandbox runs, snapshot storage is billed per GB per 30-day month while suspended, and network egress is free. Usage is pooled across your organization and only whole CUs are billed. The console shows your remaining balance, recent activity, and an estimate of usage that has not finalized yet.
Yes. Your data, projects, API keys, and remaining Usage Credits are preserved across plan changes. Unused credits are retained rather than refunded, and become available again if you return to Usage Credits. Returning to the plan also requires purchasing a new pack.
Email support@tensorlake.ai and one of our founders will reach out directly. For procurement, security reviews, or anything Enterprise, book a 30-minute call and we'll pull in the right person on our side.