Early Technical Tester Guide

Start cheap, escalate only when the proof gap is real.

Agent Research Desk is a pay-per-call trust and diligence API for AI agents. It helps agents make first-pass decisions before they contact, approve, buy from, rely on, or escalate a company, vendor, claim, or lead.

Invite source manual-dm is tracked only as an aggregate tester cohort.
Why you are being invited You are being invited because this service is already live on Base mainnet and the next job is controlled usability testing, not payment infrastructure testing.
Fastest first route Start with /api/company-intel for $0.05.
Escalate to core routes Escalate to /api/vendor-risk for $0.05 when the buyer already knows the company is interesting and now needs a sharper trust or control read.
Deep Diligence boundary Only use /api/deep-diligence for serious proof-gap mapping before a real decision.
Route advisor If you are unsure, use /start?src=manual-dm for a direct first-buy recommendation and starter payload.
Route Advisor https://agent-research-brief-api-production.up.railway.app/start?src=manual-dm Activation Path https://agent-research-brief-api-production.up.railway.app/activation?src=manual-dm Inspect Schema https://agent-research-brief-api-production.up.railway.app/docs.json?src=manual-dm Live Status https://agent-research-brief-api-production.up.railway.app/status?src=manual-dm Public Dashboard https://agent-research-brief-api-production.up.railway.app/dashboard?src=manual-dm Landing Page https://agent-research-brief-api-production.up.railway.app/?src=manual-dm

Why This Matters

What we need from controlled testers

  • You are being invited because this service is already live on Base mainnet and the next job is controlled usability testing, not payment infrastructure testing.
  • The fastest useful signal is whether the route ladder, docs, and payment handoff are clear enough for a technical buyer to finish a first purchase without help.
  • The most valuable early feedback is where you got confused, hesitated, or felt the product promise was unclear.

Fastest Test Path

How to get signal quickly

  1. Start with POST /api/opportunity-check at $0.01 to see the lightest-weight trust screen.
  2. If you already have one named company but want a better reason before paying, use GET /trust-preview or POST /api/trust-preview first.
  3. If the target still looks worth attention, escalate to POST /api/company-intel or POST /api/vendor-risk at $0.05.
  4. Only buy POST /api/deep-diligence when the decision is serious enough that missing evidence could change the outcome.
  5. Inspect /docs.json before you buy so you can see the exact request and response shape.
  6. Helpful feedback: tell us where the docs, pricing ladder, or payment handoff felt unclear.

Suggested Path For This Source

Manual direct outreach

  • Start with /api/company-intel at $0.05.
  • It shows a clear before-contact or before-approval use case without forcing the buyer to understand every route on the first pass.
  • Escalate to /api/vendor-risk at $0.05 only when the buyer already knows the company is interesting and now needs a sharper trust or control read.
  • Do not start with research-brief because the broadest route slows the buyer down when the outreach is about one named target and a simple trust decision.

First Purchase Guide

If you are choosing between Opportunity Check and Company Intel

Use the cheap screen when you are still deciding whether the target deserves more spend. Use Company Intel only when you already have one named company and want a real snapshot.

check_opportunity

/api/opportunity-check

Buy this when: I only want the cheapest first paid test.

Price: $0.01

Boundary: Do not skip this step unless you already know the exact blocker.

create_company_intel

/api/company-intel

Buy this when: I already have one named company and want a fast trust snapshot.

Price: $0.05

Boundary: If you are still unsure the target deserves more than a cheap screen, start with Opportunity Check first.

assess_vendor_risk

/api/vendor-risk

Buy this when: I already know the vendor and need an approval or rollout read.

Price: $0.05

Boundary: Do not use this as your first purchase if you only want to test the paid flow cheaply.

check_source_claim

/api/source-check

Buy this when: One claim or source is the blocker.

Price: $0.05

Boundary: Do not buy a broader route if one focused proof question is the real issue.

create_deep_diligence_report

/api/deep-diligence

Buy this when: Missing evidence could change a real decision.

Price: $0.50

Boundary: Do not use this as a first purchase unless the decision is already meaningful enough to justify the premium step.

x402 Flow

What the 402 to paid 200 path should feel like

Inspect the docs first, expect the unpaid challenge, then retry with x402 on Base mainnet and confirm you receive structured JSON.

Inspect

Read the schema first

  • Open /docs.json?src=manual-dm to inspect the request and response shape.
  • Pick the cheapest route that answers your question.
  • Use the copy-paste payloads below to remove guesswork.

Challenge

Expect the unpaid step

  • Send an unpaid POST to a protected route first.
  • A clean HTTP 402 means the paywall is working as intended.
  • This is expected friction, not a broken service failure.
  • The challenge should also remind you when to fall back to /api/opportunity-check first.

Complete

Retry with payment

  • Retry with x402 on eip155:8453.
  • A successful paid call should return HTTP 200 with structured JSON.
  • If the output is useful, it should also make the cleanest next paid move obvious.

Repeat Use

What a good second paid call looks like

The strongest signal is not just one paid response. It is a first route that naturally points to the next route without confusion.

Second Paid Call Standard

Buyer rule

  • Every paid response now includes commercial.repeatUsePlan so the buyer can see the single best next route immediately.
  • The best next paid step should stay on the same subject while the trust question is still warm.
  • If the next blocker is still unclear, the safest fallback is the cheap front door instead of overbuying the next route.

What to look for

Healthy second-step signal

  • The response makes one best next route feel obvious.
  • The next route stays on the same subject instead of restarting the work.
  • The buyer can tell whether to continue now, after proof, or after one outside action.

Copy-Paste Payloads

Use these first

These are synthetic starter payloads for the three most useful early test routes.

Opportunity Check

/api/opportunity-check

$0.01

POST /api/opportunity-check

{
  "subject": "Agent trust scoring for crypto vendors",
  "buyer": "Crypto fund operator",
  "marketType": "vendor",
  "geography": "United States",
  "goal": "Decide whether this topic is strong enough to package into a deeper diligence offer.",
  "knownSignals": [
    "Operators keep asking for a faster first-pass trust screen before pilots",
    "Funds want a lightweight way to compare agent vendors before deeper diligence"
  ],
  "constraints": [
    "Keep the first offer lightweight",
    "Do not assume the buyer already trusts agent vendors"
  ]
}

Company Intel

/api/company-intel

$0.05

POST /api/company-intel

{
  "companyName": "Alpine Foods",
  "website": "https://example.com",
  "industry": "Food distribution",
  "geography": "United States",
  "researchFocus": "growth",
  "knownClaims": [
    "Regional specialty food distributor",
    "Expanding private-label program"
  ],
  "knownSignals": [
    "Hiring for operations and sales",
    "Launching new product categories"
  ],
  "questions": [
    "How concentrated is demand around a few buyers?",
    "What trust signals matter most to partners?"
  ]
}

Vendor Risk

/api/vendor-risk

$0.05

POST /api/vendor-risk

{
  "vendorName": "Northwind AI",
  "website": "https://example.com",
  "serviceCategory": "AI workflow automation",
  "useCase": "Automate intake triage and case routing for an operations team.",
  "dataSensitivity": "high",
  "criticality": "high",
  "geography": "United States",
  "knownControls": [
    "SOC 2 Type II claimed",
    "SSO and role-based access controls"
  ],
  "knownConcerns": [
    "No published uptime history",
    "Third-party model dependency"
  ],
  "questions": [
    "What is the rollback plan if the workflow misroutes tasks?"
  ]
}

Feedback

What feedback helps most

  • Which route you tested first and why you chose it.
  • Whether the price ladder felt intuitive or whether you were unsure when to escalate.
  • Whether the 402 challenge and paid retry flow felt obvious or confusing.
  • Whether the output was clear enough to use in an agent workflow without manual rewriting.
  • The single biggest confusion point that would have made you stop or bounce.

Feedback API

POST /api/feedback

Feedback is stored privately and exposed publicly only as aggregate metrics. Wallet address is not required, and raw comments are never shown on public pages.

POST /api/feedback

Supported source codes: warm-technical, x402-builder, mcp-builder, base-builder, manual-dm

{
  "routeTested": "/api/company-intel",
  "source": "manual-dm",
  "testerType": "agent-builder",
  "clarityScore": 4,
  "outputValueScore": 4,
  "paymentFrictionScore": 3,
  "wouldUseAgain": true,
  "biggestConfusion": "I wanted one clearer hint on when to escalate from company-intel to vendor-risk.",
  "requestedImprovement": "Add one decision rule under each route card.",
  "comment": "The route was useful, but I wanted a clearer hint about when to escalate to deep diligence."
}

Privacy

Public metrics are aggregate only

Agent Research Desk does not publicly expose buyer prompts, wallet-level request history, or paid response contents. Public metrics are aggregate only.