Query your AI visibility, citations, and traffic from Claude or Cursor
The AI-Advisors MCP server brings your workspace data into the AI assistant your team already lives in. Generate an access token in Settings, paste a setup snippet into Claude Code, Claude Desktop, or Cursor, and ask plain-English questions about your AI visibility, citations, AEO audits, and AI traffic. Every tool is read-only by design - your assistant can look at your data, never change it.
Overview
Your AI search data lives in a dashboard. Your team increasingly works inside an AI assistant. The MCP server connects the two: 11 read-only tools give Claude, Cursor, or any MCP-capable client a complete read surface over your workspace - tracked websites, per-prompt AI visibility, recent engine answers, citation leaderboards, AEO audit scores and recommendations, AI traffic, Google Search Console performance, brand accuracy findings, advisor recommendations, and the weekly read. "How visible are we in AI answers this month?" becomes a question you ask, not a report you pull.
The security posture is deliberately strict. Access is a personal token that displays exactly once at creation - only a hash is stored, so a lost token can never be recovered or leaked from our side. Org and website scope are fixed by the token on the server for every call, never by what the client asks for, and there are no write tools at all: an assistant cannot create, edit, or delete prompts, campaigns, audits, settings, or tokens through the MCP server. Revoking a token cuts off access on its very next request.
Setup is a paste, not a project. Copy the ready-made snippet for Claude Code, Claude Desktop, or Cursor from Settings, drop in your token, and start asking. Each token's last activity shows in Settings, so you always know which credentials are in use and which are stale.
How it works
A complete read surface over your workspace
11 read-only tools cover the websites you track, per-prompt AI visibility, recent answers, citation leaderboards, AEO audits and recommendations, AI traffic, Google Search Console performance, brand accuracy findings, advisor recommendations, and the weekly read. Your assistant can answer from the same data your dashboard shows.
AI visibility, prompt by prompt
Ask how often ChatGPT, Claude, Gemini, and Perplexity mention and cite your brand across tracked prompts, with period-over-period change over a default 30-day window and lookbacks up to 90 days. Drill into any prompt for per-platform mention and citation flags, sentiment, cited domains, the search sub-queries the engine ran, and an excerpt of the actual answer.
Citations and audit context on demand
Pull the domains AI engines cite when answering your tracked prompts, with your own domain flagged in the ranking, plus the latest AEO audit's overall, technical, content, and authority scores and up to 15 open recommendations - all without opening the dashboard.
AI traffic next to search performance
Compare workspace AI traffic - total sessions by channel, AI referral visits, and a source comparison - against Google Search Console clicks, impressions, and top queries, so your assistant can put the AI channel in context in a single answer.
Brand accuracy and advisor guidance
Surface the latest brand accuracy check - claims AI engines made about your brand that contradict your own website or could not be verified - alongside the current recommendations from the insights, AEO, ads, and automation advisors and the latest weekly what-changed-and-why narrative.
Show-once tokens with instant revoke
Tokens display exactly once at creation and only a hash is stored, so a database leak exposes nothing usable. Revoking a token takes effect on its very next request, and Settings shows each token's last activity, so unexpected use of a stale credential is visible to admins.
Copy-paste setup for the clients you use
Ready-made setup snippets for Claude Code, Claude Desktop, and Cursor, plus compatibility with any client that supports MCP over HTTP with custom headers. Generate a token, paste the snippet, and start asking questions.
Set up
Setup is a token paste, not a project - no OAuth flow, no code on your site, nothing to host.
An org owner or admin opens Settings, finds the AI Assistant Access (MCP) row under Integrations, and clicks Generate token. Give the token a name you will recognize later, like a laptop or teammate name, and copy it from the show-once modal - it is displayed exactly once, and only a hashed version is ever stored.
Under Connect your assistant, pick your client - Claude Code, Claude Desktop, or Cursor - copy the setup snippet, and replace the placeholder with your token. Any other client that supports MCP over the Streamable HTTP transport with custom headers connects the same way.
That is the whole setup. Ask your assistant how visible your brand is in AI answers this month and it can call the AI-Advisors tools directly. Each token's last activity shows in Settings, and you can revoke any token there whenever you need to.
Takes a few minutes
Data and permissions
The MCP server is a strict read surface: every tool reads, none write, and what an assistant can see is fixed by the token on the server, never by what the client asks for.
- Websites tracked in the workspace: brand name, domain, and industry
- Per-prompt AI visibility stats with period-over-period change, and tracked prompts with their intent tags
- Recent AI engine answers per prompt: platform, mention and citation flags, sentiment, cited domains, search sub-queries, and a short answer excerpt
- Citation domain leaderboards, plus the latest AEO audit scores and open recommendations
- Workspace AI traffic (sessions by channel, AI referral visits, source comparison) and Google Search Console clicks, impressions, and top queries
- Brand accuracy findings, advisor recommendations, and the latest weekly intelligence narrative
- Nothing in your workspace data - there are no write tools
- The only write on the MCP path is the token's own last-activity timestamp, which powers the activity display in Settings
- Another organization's data - org and website scope come from the token on the server for every call, and a request for someone else's website returns not found
- Prompts, campaigns, audits, settings, or tokens - an assistant cannot create, edit, or delete anything
- CRM contact records or any customer contact data - no tool reads them
- The plaintext token after creation - only its hash is stored, and the server redacts the credential from request context
There is no OAuth flow. Authentication is a personal access token sent as a bearer header on every request, and the server grants each session a single read scope - write scopes do not exist.
Good to know
- Available on Growth and Enterprise plans; AppSumo lifetime workspaces get MCP access after stacking 3 codes.
- Each token allows 120 tool calls per hour, and a workspace can hold up to 10 active tokens.
- Lookback windows go up to 90 days per query; oversized results are truncated at 8,000 characters with an explicit flag.
- Creating, listing, and revoking tokens is owner and admin only.
- Revocation is instant: the endpoint re-authenticates every request, so a revoked token fails on its very next call.
- Requires a client that supports MCP over the Streamable HTTP transport with custom headers; the legacy SSE transport is not supported.
Frequently asked questions
What can my AI assistant read, and what can it change?
It can read your tracked websites, per-prompt AI visibility, recent AI answers, citation leaderboards, AEO audit scores and recommendations, AI traffic, Google Search Console performance, brand accuracy findings, advisor recommendations, and the weekly read. It can change nothing: there are deliberately no write tools, so prompts, campaigns, audits, settings, and tokens cannot be created, edited, or deleted through the MCP server.
Which AI assistants work with it?
Any client that supports MCP over the Streamable HTTP transport with custom headers. Ready-made setup snippets ship for Claude Code, Claude Desktop, and Cursor. The legacy SSE transport is not supported.
Which plan do I need?
Growth or Enterprise. AppSumo lifetime-deal workspaces get MCP access after stacking 3 codes. It is not included on Starter.
How are access tokens secured?
Tokens are shown exactly once at creation and only a SHA-256 hash is stored, so a database leak exposes nothing usable. Verification uses constant-time comparison, revocation is re-checked on every request, and Settings shows each token's last activity, so unexpected use of a stale token is visible to admins. If a token is lost, revoke it and generate a new one - support cannot recover it, because only the hash exists.
How long does setup take?
A few minutes. An owner or admin generates a token in Settings under Integrations, copies the setup snippet for Claude Code, Claude Desktop, or Cursor, and replaces the placeholder with the token. The assistant is connected on its next question.
What happens when I revoke a token?
Access ends immediately. The endpoint is stateless and re-authenticates every request, so an assistant using a revoked token loses access on its very next call. Your data is untouched either way, because the integration is read-only.
Are there usage limits?
Each token allows 120 tool calls per hour, a workspace can hold up to 10 active tokens, tool results are capped at 8,000 characters, and data lookbacks go up to 90 days per query.
