Skip to main content
Declaw Cloud bills in microdollars (1 USD = 1,000,000 µ$) with second-level resolution. Every Pro and Enterprise operation is metered; every account starts with free credits. There are no fixed monthly fees — you pay only for what you run.

​
Wallet model

Each account has one wallet with three separate pools that drain in a strict waterfall order:
1

Sandbox free credits

$100 grant at account creation. Consumed first by sandbox compute charges (vCPU, memory, disk). Never refilled by deposits.
2

Guardrails free credits

$200 grant at account creation. Consumed first by guardrails scan charges (PII, injection, etc.). Never refilled by deposits.
3

Paid balance

Single shared pool. Topped up by deposits. Used for both sandbox compute and guardrails charges once the respective free credits are drained.
Charges against the sandbox wallet drain sandbox_free_micros first, then fall through to balance_micros. Guardrails charges do the same with guardrails_free_micros. You can watch both drain in real time via GET /accounts/:owner_id.
The 100+100 + 200 free credits are one-time grants. They do not refill monthly or on deposit.

​
Sandbox compute pricing

Sandbox compute is metered every 30 seconds based on the sandbox’s provisioned resources (vCPUs, memory, disk). The metering tick caps at 90 seconds of billable time to bound exposure from delayed ticks.
DimensionRateEquivalent
vCPU14.0 µ$ / vCPU / sec$0.0504 / vCPU / hour
Memory4.5 µ$ / GB / sec$0.0162 / GB / hour
Disk (overlay)0.0424 µ$ / GB / sec$0.110 / GB / month

​
Example

A sandbox with 2 vCPUs, 4 GB RAM, and 10 GB disk running for 1 hour costs: 
vCPU:    2  × 14.0    × 3600 =  100,800 µ$
Memory:  4  × 4.5     × 3600 =   64,800 µ$
Disk:    10 × 0.0424  × 3600 =    1,526 µ$
                               ─────────
Total:                         ~167,126 µ$  ≈  $0.167
Charges accrue continuously as the sandbox runs. Killing or pausing the sandbox stops compute metering immediately. Resuming a paused sandbox resumes metering at the same rate.

​
Guardrails scan pricing

Each guardrails scan is billed per invocation from the sandbox’s security proxy. Regex-based scans run locally in the proxy and are free. Only scans dispatched to the ML guardrails service are metered:
ScannerRate per scanEquivalent
PII (Presidio + spaCy NER)600 µ$$0.0006
Prompt injection (Prompt-Guard-86M)600 µ$$0.0006
Toxicity (RoBERTa ONNX)400 µ$$0.0004
Code security300 µ$$0.0003
Language detection (XLM-RoBERTa)200 µ$$0.0002
Invisible text100 µ$$0.0001
Regex (local, in-proxy)0Free
Guardrails charges are metered in batches every 30 seconds. The usage endpoint returns per-scanner scan counts so you can attribute cost back to which policies were actively scanning.

​
Deposits

Paid balance is topped up by explicit deposits — there is no auto-refill. Deposit amounts are bounded per tier:
TierMin depositMax depositMonthly recharge commitment
Free$5$100None
Pro$10$5,000$100 / month
Enterprise$100$50,000$1,500 / month
Deposits go straight into balance_micros and are drawn down by either sandbox or guardrails charges once free credits are exhausted.
Pro and Enterprise require a positive wallet balance for every billable API call. If sandbox free credits + paid balance drops to zero, the API returns 402 Payment Required on the next sandbox create, command, or filesystem call until you top up. See Errors for details.

​
What’s not billed

  • Sandbox list / get / kill endpoints
  • Health and status checks
  • Template listing and snapshot metadata
  • Regex guardrails scans (always local, always free)
  • Failed or rejected requests (400 / 401 / 403 / 429 responses)
  • Paused sandbox time (compute metering stops at pause, resumes on resume)