Query your AI visibility from any agent

The LLMBrandScan MCP server lets any AI agent or client measure a brand's visibility in LLM answers over a pay-per-call, bearer-authenticated endpoint. Connect at https://llmbrandscan.com/api/mcp with a token from your account, then call five tools: two read your existing scans, three launch fresh ones.

Live engine status (2026-10-20): ChatGPT (API with web search), Gemini (grounded), Claude (API with web search) and Google AI answers (AI Overviews) are answering. Perplexity Sonar is temporarily unavailable at the provider — scans run on the engines that answer, and an engine that does not answer is not charged. This note updates by itself when an engine comes back.

The LLMBrandScan MCP server is a pay-per-call Model Context Protocol endpoint that exposes brand-visibility scanning to AI agents. Unlike every other MCP server in the category — read-only wrappers over an existing subscription — LLMBrandScan's server launches fresh, billed scans on demand: an agent calls a tool, the server reserves the scan's price from the caller's prepaid ledger, runs live queries across ChatGPT (API with web search), Gemini (grounded), Perplexity Sonar, Claude (API with web search) and Google AI answers (AI Overviews) by default (an Audit; a depth parameter selects a cheaper Quick check or a deeper Deep Audit, with Grok (live search) optional), and returns a scan id to poll for the finished report. Billing follows one rule — pay only for answers that arrive — so unused reservations return automatically. A missing or revoked bearer token is rejected before any tool runs. The server advertises five tools: two zero-spend read tools and three credit-spending scan launchers, each annotated with whether it spends credits.

The five tools

Tool	What it does	Cost
`list_scans`	List your own scans (id, brand, status).	No charge
`get_scan`	Fetch one scan by id — full report when complete, status while running. Also the poll target for the launch tools.	No charge
`check_brand_visibility`	Launch a new visibility scan across the LLM engines. Optional depth selects the preset.	100 credits by default (Audit) · depth: Quick check 9, Deep Audit 160
`compare_with_competitor`	Launch a head-to-head scan against a named competitor. Optional depth selects the preset.	100 credits by default (Audit) · depth: Quick check 9, Deep Audit 160
`get_citation_sources`	Launch a scan focused on which sources the engines cite.	9 credits (Quick check — no depth parameter)

The three credit-spending tools return a scan_id immediately and never block on the scan (about 55 minutes for the default Audit) — poll get_scan with that id to retrieve the report once it completes.

check_brand_visibility and compare_with_competitor take an optional depth ( get_citation_sources always runs a Quick check):

light — Quick check — 9 credits, about 25 minutes
standard — Audit — 100 credits, about 55 minutes (the default)
deep — Deep Audit — 160 credits, about 85 minutes

Every fresh-scan tool also takes an optional max_credits — set it to refuse the call outright, with nothing reserved, whenever that call's price would exceed it. Useful for a script or scheduled agent that expects the old Quick check default and shouldn't silently start spending Audit prices instead.

Changelog

2026-09-28 — check_brand_visibility and compare_with_competitor now default to an Audit (100 credits) instead of a Quick check (9 credits) when no depth is given. Added an optional max_credits parameter to all three fresh-scan tools — set it to keep an existing integration on its old budget.
2026-09-28 (R2-3) — fixed: max_credits is now the actual accepted tool-input field name (it was silently stripped by a camelCase mismatch in the tool schema for the first same-day release above).

Get a token

Sign in and mint a bearer token from your account page. You see the raw token once — copy it immediately and store it like a password. Every call below sends it as Authorization: Bearer YOUR_TOKEN.

Connect

1. Claude Code / MCP client config

Add the server to your MCP client's config (Claude Code, Cursor, or any streamable-HTTP MCP client):

{
  "mcpServers": {
    "llmbrandscan": {
      "type": "http",
      "url": "https://llmbrandscan.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

2. Anthropic Messages API

Pass the server in the mcp_servers array of a Messages API request, with your token in authorization_token:

{
  "mcp_servers": [
    {
      "type": "url",
      "url": "https://llmbrandscan.com/api/mcp",
      "name": "llmbrandscan",
      "authorization_token": "YOUR_TOKEN"
    }
  ]
}

3. Generic streamable-HTTP client

Point any MCP streamable-HTTP client at https://llmbrandscan.com/api/mcp and send the bearer header on every request. A raw JSON-RPC tools/list probe:

export LLMBRANDSCAN_TOKEN="<paste your token>"

curl -X POST https://llmbrandscan.com/api/mcp \
  -H "Authorization: Bearer $LLMBRANDSCAN_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Ready to start? Mint a token and call check_brand_visibility.