Verisource — B2B GEO & AI Citation Intelligence MCP

Audit ChatGPT, Gemini, and Perplexity citations, flag pricing hallucinations, and ship JSON-LD, /llms.txt, and /agents.md across 7 CMS stacks.

Hosted MCP Server

npx add-mcp 'https://mcp.verisource.dev/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Verisource — B2B Generative Engine Optimization (GEO) & Streamable HTTP /mcp Platform

AI Search Readiness — Verisource

Live Platform: https://verisource.dev · Streamable HTTP MCP: https://mcp.verisource.dev/mcp · 1-Click OAuth 2.1 Connector: https://mcp.verisource.dev/oauth/mcp · Hosted Brand /mcp Demo: https://mcp.verisource.dev/b/verisource.dev/mcp · Official MCP Registry: dev.verisource/mcp@1.0.0 · Cursor Directory: verisource-geo-ai-citation-mcp · Chrome Web Store: Install Side Panel

Verisource audits how ChatGPT Search, Google Gemini, and Perplexity cite your brand, catches hallucinated pricing and policy claims against a verified Brand Truth Sheet, and generates syntax-validated schema.org JSON-LD, /llms.txt, /agents.md, /ai-facts, and You vs. Competitor HTML comparison pages across 7 CMS platforms (Shopify, WordPress, Webflow, Wix, Squarespace, Framer, and Next.js).


Live Free Tools, Benchmarks & Sandboxes (Zero Account Required)

ResourceURLWhat It Does
Free 60-Second Domain Auditverisource.dev/#scanProbes 5 AI crawlers, 3 answer engines, 5-page link health (VerisourceScan/1.0), CMS stack, and 3-pillar JSON-LD readiness.
Public Audited Brand Directoryverisource.dev/reportsSearchable protocol leaderboard and live SVG readiness badges across 510+ audited SaaS and commerce domains.
Head-to-Head Domain X-Rayverisource.dev/vsSide-by-side 6-dimension HTTP protocol, WAF, and JSON-LD benchmark for any two public domains.
11-Capability Platform & CLI Tourverisource.dev/featuresInteractive split-screen walkthroughs for AI search verification, browser flow replay, route mapping, and log correlation.
Crawler Transparency Specverisource.dev/scannerPublic VerisourceScan/1.0 User-Agent specification, rate limits, and /robots.txt opt-out controls.
2026 40-Domain AI Readiness Studyverisource.dev/benchmarks/ai-search-readiness-2026Empirical HTTPS crawl dataset across B2B SaaS, Shopify D2C, agencies, and local brands.
4 Free Protocol Validatorsverisource.dev/toolsZero-signup checkers for 3-pillar JSON-LD & FAQPage, /llms.txt, 5-Bot WAF, and Shopify AI.
Interactive 90-Second Console Tourverisource.dev/tutorialClick-through sandbox for Business Owners, White-Label Agencies, and /mcp Developers.
Live Agency White-Label Sandboxverisource.dev/agencies#sandboxPreview your agency brand, hex accent bar, and booking URL on any client domain in 15 seconds.
7-CMS Templates & Connectorsverisource.dev/connectors · /templatesDrop-in verisource-geo-connector.php WordPress plugin, Shopify OS 2.0 Liquid bridge, and templates.
AI Conversion Payback Calculatorverisource.dev/pricing#vs-pricing-roi-calculatorModels 14.2%–15.9% AI referral buyer conversion vs. 1.76%–2.8% organic baseline across 4 verticals.
25% Recurring Partner Programverisource.dev/partners90-day attribution cookie synced directly to Stripe subscription metadata.

Connect Claude Desktop, Cursor, Windsurf, or Antigravity in 30 Seconds (/mcp)

Verisource runs a stateless Streamable HTTP MCP server built on @modelcontextprotocol/server v2 at https://mcp.verisource.dev/mcp (published on registry.modelcontextprotocol.io as dev.verisource/mcp@1.0.0, zero deprecated /sse routes).

Add this entry to claude_desktop_config.json, .cursor/mcp.json, or your MCP client settings:

{
  "mcpServers": {
    "verisource": {
      "url": "https://mcp.verisource.dev/mcp",
      "headers": {
        "Authorization": "Bearer vs_live_YOUR_API_KEY"
      }
    }
  }
}

1-Click Claude Web (claude.ai) & ChatGPT Connectors (Zero API Keys): In Claude (Settings → Connectors → Add Custom Connector) or ChatGPT Connectors, paste https://mcp.verisource.dev/oauth/mcp and approve in 1 click via OAuth 2.1 + PKCE S256. Zero-Key Preview: Omit headers on https://mcp.verisource.dev/mcp to invoke scan_domain_ai_readiness immediately for 0 credits (3 custom domains/IP/day, 24-hour edge KV cache). Mint a vs_live_ SHA-256 Bearer key on Pro ($99/mo · 1,000 credits/mo, $0.00 7-day trial) or Agency ($249/mo · 5,000 credits/mo) at verisource.dev/pricing.

Exposed MCP Tools

  1. scan_domain_ai_readiness (Free · 0 credits): Returns AI Visibility Score (0–100), 7-platform CMS fingerprint, 3-pillar JSON-LD inspection (Entity Identity, FAQPage Q&A count, Offer pricing), 5-bot WAF & /robots.txt status, Shopify /api/ucp/mcp + /agents.md classification, and sample buyer prompt citation check.
  2. get_citation_gaps_and_fanouts (Pro / Agency · 1 cached / 10 live credits): Returns Mention vs. Citation rates across ChatGPT, Gemini, and Perplexity, Source-Type categorized citation gaps (Reddit, G2, Clutch, editorial media), and observed search fanout sub-queries.
  3. check_brand_hallucinations (Pro / Agency · 1 cached / 10 live credits): Diffs multi-engine claims per bullet against your verified Brand Truth Sheet so rival prices never trigger false alerts.
  4. generate_geo_fix_bundle (Pro / Agency · 1 cached / 10 live credits): Outputs syntax-validated JSON-LD, /llms.txt, /agents.md, /ai-facts, Shopify templates/agents.md.liquid, and 1-Click AI Remediation Prompts.
  5. compare_competitor_protocol_xray (Pro / Agency · 1 cached / 10 live credits): Benchmarks own_domain vs. rival_domain across 6 HTTP & schema dimensions and generates an overtake prompt + [Brand] vs. [Rival] HTML page.

Hosted Customer Brand /mcp + W3C Chrome WebMCP (navigator.modelContext) Infrastructure

Every Starter ($49/mo), Pro ($99/mo), and Agency ($249/mo) domain automatically receives a hosted, read-only Brand Truth MCP Server (https://mcp.verisource.dev/b/{domain}/mcp) plus an edge-cached W3C Chrome WebMCP script (https://mcp.verisource.dev/b/{domain}/webmcp.js) exposing get_verified_pricing_and_offers, verify_brand_claim, compare_with_competitor, and get_official_faqs from a 2-line <head> embed—zero backend code required.


Local-First CLI & Deterministic Browser Flow Verifier (@verisource/cli)

Run AI readiness checks, 20-page link health crawls, static Next.js/React route mapping, and deterministic .verisource/workflows/*.json replays from any terminal or CI runner on Node.js 20+ (zero runtime sub-dependencies):

# Audit AI crawler WAF rules, JSON-LD, /llms.txt, /agents.md & 20-page site health
npx @verisource/cli scan https://verisource.dev

# Enforce CI policy checks (.verisource/checks.json -> exit 0 pass, 1 fail, 2 error)
npx @verisource/cli verify https://verisource.dev --init

# Map static Next.js / React routes + JSON-LD emitters or install coding agent skills
npx @verisource/cli map . --diff
npx @verisource/cli agent --skill --install all

# Direct zero-auth HTTPS tarball execution (works anywhere before npm login)
npx https://verisource.dev/downloads/verisource-cli.tgz --help

Key Empirical Findings (2026 40-Domain HTTPS Benchmark)

Direct HTTP probes across 40 commercial domains (full dataset):

  • Silent Edge WAF HTTP 403 Blocks: 0/40 disallowed GPTBot in /robots.txt, yet 5.0% returned HTTP 403 Forbidden at the CDN edge when probed with real GPTBot/1.2 or PerplexityBot/1.0 headers.
  • SPA Soft-404s & Unedited /llms.txt Boilerplate: 30/40 returned status 200 on GET /llms.txt, but 25.0% were either client-side text/html soft-404 shells (7.5%) or unedited plugin dumps (17.5%). Only 50.0% serve curated Markdown specifications.
  • 80-Point Shopify Agent Truth Gap: 80.0% of audited Shopify D2C stores respond on /api/ucp/mcp, yet 0.0% customized /agents.md with real shipping thresholds or return windows (5.0% across all 40 brands).
  • Missing Answer-First FAQPage Schema: 82.5% ship base JSON-LD, while only 7.5% include FAQPage Q&A nodes.

Monorepo Architecture

Workspace PackagePurposeStack
packages/core-geoWinterCG-portable intelligence engine: hop-validated SSRF guard, 5-bot WAF prober, VerisourceScan/1.0 site health crawler, 7-CMS detector, 3-pillar JSON-LD inspector, Shopify /api/ucp/mcp adapter, Hallucination Watchdog, and vs_live_ SHA-256 key utilities.TypeScript 5.7, Zod, Web Crypto
packages/cliStandalone esbuild-bundled @verisource/cli binary: scan, verify, show, query, map, browse, record, run, workflow check, log, and agent --skill --install.Node.js 20+ ESM
packages/uiShared Obsidian Slate & Warm Brass design system (.vs-audit-paper-surface light-paper report isolation): ThreeTabShell, CopyCommandPill, LightBrowserFrame, ConsoleQuickJumpBar, CodeArtifactCard, MetricLedgerCell, StatusBadge.React 19
apps/web8-Band Dark Studio Homepage (/), 11-Capability /features & /scanner, 3-Step Subscriber Console (/dashboard), White-Label Agency Scorecards (/audit/[slug]), Free Validators (/tools/*), Origin/HMAC Pixel (/api/pixel), and Stripe Billing.Next.js 15 App Router, Supabase, Stripe
apps/mcp-workerStateless Streamable HTTP /mcp server (createMcpHandler), RFC 9728 Protected Resource metadata, KV-cached SHA-256 Bearer + OAuth 2.0 auth, atomic credit metering, and scheduled cron scanner.Cloudflare Workers (nodejs_compat)
apps/extensionLeast-privilege (['sidePanel', 'activeTab'] only) Chrome Side Panel sharing packages/ui and packages/core-geo.WXT 0.19, Manifest V3

Local Development & Verification Suite

Out of the box, PROVIDER_MODE=mock runs real outbound HTTPS/WAF probes while simulating LLM answer snapshots, Supabase state, and Stripe checkout locally with zero external API keys required.

# Install workspace dependencies
pnpm install

# Start the Next.js Web Platform & 3-Step Console (http://localhost:3000)
pnpm dev:web

# Start the Cloudflare Streamable HTTP /mcp Worker (http://localhost:8787/mcp)
pnpm dev:mcp

# Run all unit, /mcp Worker, production build, and live HTTP E2E suites
pnpm typecheck
pnpm test
pnpm build
pnpm --filter @verisource/web test:e2e

Discovery & Protocol Endpoints