CiteGate is live: one x402 tool that checks whether AI engines can cite your pages — and whether an MCP surface is safe for an agent to pay through. It ships with three tools today, priced in USDC on Base: a deterministic citability audit ($0.01), a robots-policy check ($0.01), and a live MCP gate test ($0.25). Product page: https://scriptmasterlabs.com/citegate
Why one tool
ScriptMasterLabs spent October measuring two refusals. Citation refusal: AI engines answer questions in our lane without citing the sources they could fetch. Payment refusal — the healthy kind: a decision gate holding an agent's money back from an MCP surface whose destructive tools are exposed. Cite Me Live (https://scriptmasterlabs.com/cite-me-live) published the policy proof: block training crawlers, allow every answer crawler, get cited anyway — 188 Microsoft Copilot citations at the last Bing Webmaster reading (Oct 15, window Jul 14–Oct 13). The MCP Safety Index (https://scriptmasterlabs.com/mcp-safety-index) published the gate proof: 7 vendor MCP surfaces gate-tested in 7 days. CiteGate wraps both measurement systems into tools anyone can call.
The tools
audit_citability ($0.01) — deterministic. Fetches your page and its robots.txt, then scores 0–100 against a printed rubric: HTTP 200 (15), self-referential canonical (10), title (10), meta description (10), JSON-LD present (15), a 40–60 word self-contained answer block (20), answer crawlers allowed (10), training crawlers blocked (5), no wallet address sitting in liftable page text (5). Every failed check returns a plain-language fix. First live subject: our own /about — 100/100, answer block 49 words.
robots_policy ($0.01) — deterministic. Returns your domain's crawler split (training vs answer/search, blocked or allowed, from the actual robots.txt) plus a ready-to-paste policy snippet implementing the pattern: GPTBot, CCBot, ClaudeBot and Google-Extended blocked; OAI-SearchBot, ChatGPT-User, Claude-SearchBot, PerplexityBot, Googlebot and Bingbot allowed. On scriptmasterlabs.com it reads 4/4 training blocked, 6/6 answer allowed.
gate_test_mcp ($0.25) — the same harness behind the Safety Index (decider local-heuristic-v1, calibrated=false, printed on every result). Counts a surface's capabilities from its public tools list, then runs the three canonical scenarios — bounded read, confirmed destructive, no-ask destructive — against the decision bands: >=0.80 auto-act, 0.50–0.79 hold for review, <0.50 escalate.
The receipt
First paid call, run as an owner self-test from our proof wallet (not a customer purchase — labeled so our numbers stay honest): $0.01 audit of /about, settled on Base in block 52,802,634, tx 0x5635a60e55f33feb03889c04432dc818522219ef11d5a5b334773d1021bf6be4 — https://basescan.org/tx/0x5635a60e55f33feb03889c04432dc818522219ef11d5a5b334773d1021bf6be4
The honest ceiling
There is no citation_check tool in v1. No engine sells citation data, so a citation probe is best-effort observation, not an API — we run ours twice a week and publish the log at https://scriptmasterlabs.com/cite-ledger rather than sell a number we can't guarantee. The deterministic tools are exact. The gate decider is a heuristic and says so on every result. Endpoints: GET https://squeezeos-api.onrender.com/citegate/ (free info), with citegate.scriptmasterlabs.com attached and resolving as soon as its DNS record lands.
Top comments (0)