We run the pipeline. You get the report.
A premium tier for the moments quality matters more than per-call cost. We run Scout, Forge, or Prove on top-tier LLM (Claude Opus 4.7 / GPT-5.5 Pro), a human reviewer audits every claim, and you get a polished deliverable in 24-72 hours.
Two reasons to pick this over the self-serve product: (1) you don't want to manage your own LLM API key, and (2)you're shipping the report to a customer, investor, or board — quality of the final document matters more than the per-call price.
When to use Done-For-You vs self-serve
| Scenario | Pick |
|---|---|
| You have an LLM API key and want to iterate cheaply within your annual quota | Self-serve (BYOK) |
| You're an AI agent calling our API at machine speed | Agent API ($0.05-$15/call) |
| You want a polished report you can send to a customer | Done-For-You |
| You don't want to manage your own API key | Done-For-You |
| You need maximum quality for an investor pitch | Done-For-You |
| You've hit your annual quota and need more runs this cycle | Done-For-You (overage) |
The three services
Each service is sold separately — only pay for the stage you need. You can buy Scout alone for $39, then upgrade to Forge / Prove later if the gaps look promising.
Scout Run
We scan your target sectors, ingest fresh signals, score and cluster pain points, then synthesize a complete market gap report — all on top-tier LLM with a human pass for fact-checking and polish.
Forge Run
We run the 5-round multi-agent ideation against your gaps (Scout output or your own brief), with all 10 agents firing and a manual screening pass on top of the automated kill votes / RICE scores.
Prove Run
Our most demanding pipeline — 10 agents debate your idea adversarially across multiple rounds, with Phase A5 fact-checking every claim against cited URLs and a human reviewer adjudicating before delivery.
What makes it different from self-serve
- Top-tier LLM stack — Claude Opus 4.7 and GPT-5.5 Pro for every agent call (not the cost-balanced Sonnet / MiniMax / GPT-5.4 we run on the self-serve and Agent API tiers).
- Human review on every report — a real person reads the LLM output, checks the cited URLs, removes hallucinated stats, tightens the execution plan.
- Polished deliverable — clean PDF with table of contents and formatted tables, plus the interactive web view for sharing a link.
- Direct line to us— reply to the delivery email and you're talking to the same human who reviewed the report.
- Refund on failed runs — if the model errors out or the human reviewer flags the output as unfit to ship, we re-run at no cost or refund per Terms § 4.
How to order
Two paths:
- From the pricing page— scroll to the "Done-For-You" section on /pricing, pick a service, pay with USDC or card.
- By email — gapsmith@draftlabs.org with the service tier and a 1-2 sentence brief (sectors / idea / target market). We'll send the invoice and start within the same business day.
Brief format (Forge / Prove)
For Forge and Prove, the more focused the brief, the sharper the output. Aim for ~150-300 words covering:
- Sector / market — concrete, e.g. "cross-channel e-commerce SMBs $500K-$10M GMV".
- Pain or hypothesis — what gap are you exploring?
- Constraints — geo / regulation / team size / budget cap.
- What you want out — "ship-ready execution plan" vs "3 alternatives to compare" etc.
If you skip the brief and just give us "find me a startup idea", we'll run the Scout pipeline first and recommend a focus area before invoicing for Forge / Prove. That keeps the spend honest.
Under the hood
Done-For-You runs on top of the same FastAPI engine as the self-serve product — but we deliver through a different surface: our internal CLI tool (the same one we ship at the $34.90 power-user tier).
The CLI gives our reviewers things the web product doesn't: batch sector scanning, scripted brief templates, parallel pipeline runs, and direct access to intermediate transcripts so a human can intervene mid-run instead of waiting for the final output. Every Done-For-You report goes through:
- CLI dispatch— engine pinned to top-tier model (Opus 4.7 / GPT-5.5 Pro) with our reviewer's reasoning notes attached to the brief.
- Reviewer-in-the-loop checkpoints — at the end of each round our reviewer reads the transcript, can re-run a stage with a sharper prompt, or inject domain context the model missed.
- Human review pass — every cited URL is opened, every hard stat is checked against source, and the final document is rewritten for clarity before delivery.
- Polished output — exported to PDF with table of contents + formatted tables, plus an interactive web view for sharing.
That's why turnaround is 24-72 hours rather than 30 minutes. The engine work is the same; the human + CLI workflow on top is what you're paying for.
Ready to order?
Pricing card on /pricing accepts USDC or card. Or email us with your brief.