TRUTH ENGINE · MCP · API

Verdicts as a service.
For agents.

Built for agents. Grounded in human voices. Verdicts you can audit. GYIBB turns 200,000+ user voices across Reddit, Hacker News, YouTube and more into structured product verdicts — with citations, sample sizes and confidence tiers on every answer.

Pricing → Add to your AI in 10s Full API docs

A verdict, not links

search engines return pages; GYIBB returns the answer — rating, recommendation state, pros/cons with the user quotes behind them.

Metered on truth

a "call" is a tools/call that returns a verdict. Handshakes and listings are free. Quotas reset monthly.

One endpoint, every agent

MCP over Streamable HTTP at gyibb.com/mcp. Claude, Codex, OpenCode and 180+ clients already connect.

Ask it something. Right now.

This is the real engine — verify_live over MCP, live scan included. No signup, no key.

Pricing.

Free

$0 / forever

  • →25 verified answers / day (anonymous)
  • →100 / day with the public integration key
  • →All four MCP tools
  • →Citations + confidence tier on every verdict
Connect — no signup

Maker

popular

$19 / month

  • →5,000 tools/call / month, pooled
  • →Issued API key + usage metering
  • →No anonymous daily caps
  • →verify_live early access (claim → live crawl verdict)
Get your key — email [email protected]

Team

$99 / month

  • →50,000 tools/call / month
  • →Everything in Maker
  • →Corpus export / licensing conversations
  • →Direct support channel
Talk to us — [email protected]

Billing is manual while we're small: pay or email, we issue your API key, usage metering and quota enforcement activate against it within the hour. Unkeyed traffic stays free and unmetered-for-now — existing agents are never cut off.

Connect in one block.

{
  "mcpServers": {
    "gyibb": {
      "url": "https://gyibb.com/mcp",
      "note": "free tier needs no key; paid plans send X-API-Key"
    }
  }
}

Paid plans: send your key as X-API-Key. When the monthly quota is reached the engine answers 429 with an upgrade pointer — never silently.

Why agents can trust it.

Every verdict is auditable

Responses carry sample sizes, per-platform source counts and a confidence tier. If we cannot defend a verdict with voices, we return insufficient_data — not a guess.

Read the methodology →

No pay-to-play. Ever.

Brands cannot buy a rating, a ranking or a verdict change. Affiliate revenue only ever comes from referral clicks on products we already recommend — and pages we don't recommend carry zero buy links.

Read the manifesto →

Open method, closed corpus

The scoring standards are versioned and published; every response names the version it was scored with. The data layer lives behind the API — that is the product.

How scoring works →

FOR BRANDS, TEAMS & CREATORS

Point your AI at your own product.

No dashboard to buy. Connect your agent and ask "what do people actually say about my product?" — the brand_pulse tool returns voice counts, platform spread, sentiment split, recurring complaint themes with frequencies and the representative quotes behind them. Any product, reviewed or not. Try it live, no signup →

The mirror is read-only. You cannot buy, edit or suppress a rating — that's the product. → the contract