Model Context Protocol
GeoMaestros for Claude Desktop, Cursor & ChatGPT Desktop
Connect any MCP-compatible AI client to your workspace. Ask plain English questions about your AI search visibility, competitors, and tracked properties. Read-only by default. Revoke any time.
What you can ask
- "What is my share of voice this week?"
- "List my tracked hotels."
- "Which competitors should I worry about?"
- "Show me the most recent AI search evidence for my brand."
- "What are last week's biggest visibility gaps?"
Install in 60 seconds
- 1. Sign in at app.geomaestros.com.
- 2. Open Settings → MCP tokens and click Mint new token. Copy the token immediately — it is shown only once.
- 3. Open Claude Desktop's config file (macOS:
~/Library/Application Support/Claude/claude_desktop_config.json) and paste:{ "mcpServers": { "geomaestros": { "transport": { "type": "streamable-http", "url": "https://api.geomaestros.com/v1/mcp", "headers": { "Authorization": "Bearer YOUR_TOKEN" } } } } } - 4. Restart Claude Desktop. Type list my GeoMaestros properties. You should see the live answer.
7 read-only tools (write tools opt-in)
All tools are scoped to the workspace your token was minted for. Add mcp:write at mint time only if you need write actions.
list_propertiesTracked hotels, destinations, brands.list_keywordsActive queries, intent, last visibility.get_sovOverall + AI Overview + Local Pack SOV.get_share_of_voice_trendWeekly trend, last N weeks.get_evidenceRecent AI search evidence rows.get_competitorsTracked competitors with prospect score.get_visibility_gapsRecent why-they-won analyses.
Security
- MCP tokens carry an audience claim. They cannot be replayed against any non-MCP endpoint.
- Rate-limited per workspace and per user. Write tools have a tighter bucket so a leaked write token cannot burn your tracking quota in one burst.
- 90-day default expiry. Revocation propagates within 30 seconds.
- Every tool call is audited. Workspace members can read the invocation log for transparency.
FAQ
What is MCP?
Model Context Protocol. An open standard from Anthropic that lets AI clients (Claude Desktop, Cursor, ChatGPT Desktop) call external tools. Think OAuth + REST + JSON-RPC, designed for AI.
Is this secure?
Yes. MCP tokens carry an audience claim that scopes them to the MCP server only. A leaked token cannot be replayed against your dashboard, billing, or any other endpoint. The signing key is independent of the web session secret.
How do I revoke a token?
Open the MCP tokens page in your dashboard settings and click Revoke. Within 30 seconds the token is invalidated everywhere.
How long do tokens last?
90 days by default. You can revoke earlier or mint shorter-lived tokens via the API.
Why does write access need a separate scope?
Read tools (SOV, evidence, competitors) need only mcp:tools. Mutating tools that enqueue tracking runs need mcp:write. We default to read-only so a leaked token cannot burn your tracking quota.