Quickstart

From zero to first report in 5 minutes.

There are two ways to use GapSmith: as a human via the webapp, or as an autonomous agent via the Agent API. This page covers both.

Takeaway

Buy Scout($4.90 lifetime), drop in your LLM API key, pick 3 sectors, and you'll have your first market gap report in ~6 minutes for ~$1.50 of model spend (your key, your cost).

For humans (web app)

1. Create an account

Sign up with Google OAuth or email at /signup. Email is auto-verified for Google sign-in; magic-link is sent for email signup.

2. Buy a tier

Pricing follows a bonding curve — earlier buyers pay less. Two payment rails:

  • USDC on Solana (primary, recommended) — Click Pay 4.90 USDC, approve in Phantom, ~3 second settlement.
  • Card via Stripe — same price, demoted to a text link.

Each purchase comes with an annual usage quota (12 Scout runs / 6 Forge sessions / 4 Prove debates) that resets every 365 days from the purchase date. Run out? Order a Done-For-You report instead — Claude Opus + human-reviewed.

3. Add your LLM API key

GapSmith is BYOK — your API costs go directly to your provider, not to us. Drop a key into /settings. Anthropic, OpenAI, Google Gemini, and MiniMax are supported. You can save one key per provider — the run page picks the right one based on the model you choose.

4. Run Scout

Pick up to 10 industry sectors (AI/ML, SaaS, Fintech, etc.), pick a model (start with claude-sonnet-4-6 or MiniMax-M2.7 for the balanced default), and click Get Scout. Progress streams to the page in real time. Output:

  • Daily brief synthesizing 70+ articles + 250+ pain signals
  • 3 venture-grade topics with concrete wedges + competitor pricing
  • 10 pain clusters mapped to sector + frequency
  • 10 cross-signals linking news → pain → startup opportunity

5. (Optional) Pipe into Forge & Prove

On the Scout report page, click Forge this report to run 5-round multi-agent ideation against the gaps. Or hand-pick an idea and click Prove for a multi-round adversarial debate (up to 4 rounds) with kill votes, RICE scoring, and pivot detection.

For agents (programmatic)

AI agents call our endpoints under x402 — pay-per-request in USDC, no API key, no signup. Quick smoke-test against devnet:

bash
# Install x402 client deps
pip install solders solana requests base58

# Point at your devnet wallet (or request a test wallet from us)
export AGENT_WALLET_SECRET="<your-base58-secret>"

# Run the reference demo — pays ~$0.55 devnet USDC across 5 endpoints
python agent_demo.py --skip-compute

The full reference implementation is ~80 lines — request a copy via /contact if you'd like to skip writing the x402 client yourself.

Each endpoint follows the standard x402 dance: probe (no payment) → 402 with payment requirements → sign Solana SPL transferChecked → resubmit with X-Payment header → 200 with JSON. Same code path on mainnet.

Minimum 3-line client

python
from gapsmith_x402 import probe_pay_get  # ~80 lines, see examples/agent_demo.py

resp = probe_pay_get(
    "https://gapsmith.draftlabs.org/api/v1/scout/gaps?sector=ai-ml",
    wallet_secret=os.environ["AGENT_WALLET_SECRET"],
    network="mainnet",
)
print(resp.json()["gaps"][:3])

What to read next

  • Architecture — system overview, where each piece runs, how data flows
  • Pipelines — what Scout / Forge / Prove actually do, with cost + quality benchmarks
  • x402 on Solana — how the payment protocol works, why we picked it, idempotency design
  • Agent API reference — every endpoint, parameter, error code
← Previous
Overview
Next →
Architecture