Pricing & Compute Units
The four tiers, the CU model, and what a typical run costs.
Blueprint is priced by Compute Units (CU). Every plan ships a monthly CU allowance; usage above the allowance can be covered with Top-Ups or by upgrading.
The four tiers
| Plan | Price | Monthly CU | Carry cap (rollover max) |
|---|---|---|---|
| Free | $0 | 50 | 100 |
| Starter | $29 / mo | 500 | 1,000 |
| Pro | $149 / mo | 3,300 | 6,600 |
| Team | $749 / mo | 16,500 | 33,000 |
Plans renew monthly. You can upgrade, downgrade, or cancel anytime from the Billing page.
What is a Compute Unit?
A Compute Unit is the metering unit Blueprint uses. 1 CU ≈ $0.01 of the underlying compute cost (LLM tokens + retrieval + persistence).
The CU model abstracts over the messy reality of LLM pricing: token counts vary per provider, prompt vs response tokens are priced differently, retrieval has its own cost. CU is what you actually budget against.
How much CU does a run consume?
For a typical Phase-1.4 run:
| Brief length | Approx CU |
|---|---|
| 500 words (short) | ~25 CU |
| 1,500 words (typical) | ~42 CU |
| 3,000 words (detailed) | ~70 CU |
| 10,000 words (long) | ~180 CU |
These are estimates. The pre-run estimate uses the median of similar Briefs.
Phase affects cost
Lower phases stop the pipeline earlier and cost less:
| Phase | Stops after | Approx CU (1,500-word Brief) |
|---|---|---|
| 1.1 | Stakeholder analysis | ~12 CU |
| 1.2 | Requirements | ~22 CU |
| 1.3 | Decomposition | ~30 CU |
| 1.4 | Full bundle including phase gates | ~42 CU |
Most production work uses Phase 1.4.
Verifier judges affect cost
The number of internal critic-loop judges scales accuracy and cost:
- 1 judge — cheapest, lowest assurance
- 4 judges — recommended for production runs (~1.3× cost)
- 8 judges — highest assurance, ~1.7× cost
How is CU charged?
CU is reserved at submit and finalized at completion:
- You submit; the estimated CU is reserved from your subscription balance
- The run executes, consuming actual CU
- At completion: the actual CU is debited; the difference vs reserve is refunded (if you used less) or paid from your top-up balance (if you used more)
- If a run fails for a platform reason: the entire reservation is refunded automatically