// choose your review capacity
Build on a stronger signal.
Build credits fund a complete committee build: candidate specs, synthesis, cross-provider review, deterministic structural validation, and reviewer-derived scoring.
Credits pay for the review process, not a guaranteed outcome. You receive the resulting spec and its evidence even when reviewers identify unresolved work.
MOST POPULAR
Pro
// production-ready specs
$
29
/mo
billed monthly · cancel anytime
- 8 build credits/month — one credit = one full committee build. Unused credits carry over for 90 days
- Agentic MCP access — commission specs from Claude Code, Cursor, Codex & Grok
- Full review pipeline — up to 8 rounds of cross-provider frontier review + reviewer-derived rubric score
- GitHub push-to-private-repo
- Post-run polish: 1 full-spec pass + 2 targeted passes
- Spec interview Q&A
- Email support
Start Pro plan →
Team
// scale your spec output
$
129
/mo
billed monthly · cancel anytime
- 36 pooled build credits/month — team pooling coming soon
- Everything in Pro
- Deepest-model reviewer on every build*
- API access (X-Agent-Api-Key)
- Priority support
Team pooling coming soon
Enterprise
// for organizations + platforms
Contact us
annual contract
- Custom build-credit volume
- Everything in Team
- API access (X-Agent-Api-Key)
- Custom contracts & invoicing
- Direct founder support
Contact us →
Full feature comparison
| Feature |
Pro |
Team |
Enterprise |
| Build credits per month |
8 |
36 (pooled) |
Custom |
| Reviewer-derived rubric score |
✓ |
✓ |
✓ |
| Structural validator (⚙) |
✓ |
✓ |
✓ |
| Cross-model builder panel (3 personas) |
✓ |
✓ |
✓ |
| Deepest-model reviewer |
— |
✓ |
✓ |
| Cross-model reviewer audit loop* |
standard |
premium |
premium |
| Pre-spec interview Q&A |
✓ |
✓ |
✓ |
| Committee review rounds |
up to 8 |
up to 8 |
up to 8 |
| Post-run polish passes |
1 full + 2 targeted |
1 full + 4 targeted |
1 full + 6 targeted |
| GitHub push-to-repo |
✓ |
✓ |
✓ |
| Agentic MCP access (Claude Code / Cursor / Codex / Grok) |
✓ |
✓ |
✓ |
| API access (programmatic) |
— |
✓† |
✓ |
| Support |
Email |
Priority |
Direct |
* Two distinct cross-model mechanisms: the builder panel — three independent reviewer personas on different model families (xAI, Anthropic, OpenAI) — scores every paid spec; the reviewer audit loop — two reviewers (Anthropic + OpenAI) that revise and re-review until both sign off, up to 8 rounds per build — runs on every paid tier (Pro on the standard reviewer model; Team and Enterprise on the premium model). Both use the configured provider keys.
† API access on Team is provisioned on request during early access — email support@sparkvibeai.com after subscribing and we'll issue your key.
Agentic MCP access (Pro and up): generate a key from the account page and add the SparkVibe MCP server (https://api.sparkvibeai.com/mcp) to Claude Code, Cursor, Codex, or Grok.
Post-run polish passes: after your spec is delivered, a full pass rewrites the whole spec; each targeted pass rewrites its weakest section — more targeted passes on higher tiers. Separate from the committee's review rounds, which run during every build.
Frequently asked questions
What happens to specs when I cancel?
You keep every spec you've ever commissioned, and you keep every build credit you've already paid for. Credits stay usable after you cancel until their normal 90-day expiry, so you can log in next month and spend what you bought. Export your spec data anytime via the dashboard (markdown export). Cancellation only stops recurring billing. Your agent (MCP) keeps working for as long as you have credits.
What counts as a "committee run"?
One committee run = one work request you commission, which produces 4 candidate specs (planner, security, buildability, synthesizer), the synthesized winner, the full review pipeline (dual-reviewer audit, structural validation, and reviewer-derived score), and an auto-polish pass. Re-running the committee on an existing work request counts as a separate run; the optional manual polish (Pro and up) does not.
What happens when my build credits run out?
New committee builds pause until your next paid billing period grants fresh credits. There's never surprise overage billing — the product stops, it doesn't meter. Top-up credit packs are coming soon.
Do unused credits roll over?
Yes, while you're subscribed. Each month's credits stay good for 90 days from the day they're granted, and builds always spend your oldest credits first. Bank a quiet month, spend it on a busy one. Credits older than 90 days expire, like a gift certificate. Cancelling doesn't forfeit them: paid credits stay spendable until they expire.
Do I own the specs my committee writes?
Yes. You own every spec you commission. SparkVibeAI retains no rights in your spec content. We retain rights only in the Platform itself (the scoring infrastructure, rubric, software). See the
Terms of Service §5 for the full IP terms.
Why multiple independent signals instead of one score?
A single signal — no matter how good — has the same blind spots on every spec. We chain three mechanisms that fail differently: two independent reviewers on different model families (Anthropic + OpenAI) that revise and re-review until both sign off, a deterministic structural validator (no LLM), and a reviewer-derived score that reports its confidence and withholds the number when the signal is too thin. When they agree, you have confidence. When they diverge, the divergence points to exactly where the spec is weakest.
More on the moat →
What's "cross-model" actually mean on the Team tier?
The dual-reviewer audit runs the two reviewers on different model families — Anthropic Claude and OpenAI GPT — so agreement is stronger signal than two calls to the same model. Cross-model review runs on every paid tier today; Team and Enterprise get the premium reviewer seat on every build.
Can I push specs to a private GitHub repo?
Yes, on Pro and above. Connect your GitHub account from the Account page; the Push to my GitHub button on any accepted spec creates a fresh private repo in your account with SPEC.md, README, .gitignore, and empty placeholder files for paths the spec mentions. No code is generated — the scaffold is structure, the build is yours.
Can my coding agent use SparkVibe directly?
Yes — on Pro and above. SparkVibeAI ships an MCP server: add
https://api.sparkvibeai.com/mcp to Claude Code, Cursor, Codex, or Grok with a key generated from your
account, and your agent runs the whole loop —
commission_spec →
get_commission →
get_spec →
push_to_github — without leaving the editor. Committee calls consume the same build-credit balance as the web app.
What's your refund policy?
All sales are final. A committee run incurs real compute cost the moment you commission a spec, and specs are delivered immediately as digital goods, so fees are non-refundable. At checkout you confirm you want immediate access and waive any statutory cancellation or "cooling-off" period that would otherwise apply. You can cancel anytime to stop future billing — cancellation takes effect at the end of the current period, and you keep every spec you've already commissioned.
Which plans can I buy today?
Pro is available now for self-service monthly signup. Team pooling is still in development and is not available for purchase yet. For higher-volume or organization needs, email
support@sparkvibeai.com.