Vercatus in Claude and ChatGPT (MCP)
Vercatus has an MCP server, so an assistant that speaks the Model Context Protocol (Claude Desktop, Claude Code, ChatGPT and others) can read the workspace and act on it: “which engines cite us least this month”, “what does ChatGPT cite instead of us for pricing questions”, “mark the robots.txt finding planned”. MCP comes with Growth and the plans above it.
It authenticates with an API key. Every tool call is an API request with that key, so the key’s scopes, sites, addresses and rate limit apply. Make a key for the assistant:
- read for questions only; add write to let it change finding statuses and add prompts;
- limited to the sites the assistant should see.
Connect
Section titled “Connect”The server is at https://app.vercatus.com/mcp (Streamable HTTP). For clients that start a local process
there is also a stdio server, vercatus-mcp, in the command-line client, which is provided on request.
Claude Code
Section titled “Claude Code”claude mcp add --transport http vercatus https://app.vercatus.com/mcp \ --header "Authorization: Bearer ca_…"Claude Desktop
Section titled “Claude Desktop”With the command-line client installed, add the stdio server in claude_desktop_config.json (Settings, Developer, Edit config):
{ "mcpServers": { "vercatus": { "command": "vercatus-mcp", "env": { "VERCATUS_API_KEY": "ca_…" } } }}Restart Claude Desktop; the tools appear under the tools menu.
ChatGPT
Section titled “ChatGPT”Settings, Apps & Connectors, Advanced settings: turn on Developer mode, then Create a connector with
the URL https://app.vercatus.com/mcp and API-key authentication, pasting the key. Workspace admins of
ChatGPT Business and Enterprise decide whether members may add custom connectors.
Other clients
Section titled “Other clients”Any MCP client that sends a bearer token can use the HTTP endpoint: POST JSON-RPC to /mcp with
Authorization: Bearer ca_…. The endpoint is stateless (no session id); each request stands alone.
The tools read the site audit, its own engine checks and the prompts your workspace saved. Topics, Answer Share, plays and verdicts are in the app and in the API.
| Tool | What it answers |
|---|---|
list_sites |
The sites the key can see, with their ids, verification and competitors |
get_visibility |
Per engine for a period, over the run checks and saved prompts: how often answers cited the site, with the interval, named without a link, questions asked and cited, the change against the previous period (“no clear change” when it is within the interval), crawler access, and the domains cited instead |
get_competitors |
Engine × domain for one site, over the same answers: how often the site, its named competitors and the domains cited most instead were cited |
list_findings |
Site audit findings from each site’s latest run, filtered by site, severity, status or category |
get_finding |
One finding in full: summary, recommendation, evidence, the knowledge-base entries it rests on, status and plan |
update_finding_status |
Set Open, Planned, Done – checking, Not doing or Not a problem, with a note (write scope). Fixed is set by a later run |
list_prompts |
The prompts the workspace saved, with each engine’s latest answer: cited, named or neither |
track_prompt |
Save a new prompt on a site (write scope); it also joins the topic it fits |
get_prompt_history |
A saved prompt’s answers per engine, what changed and the sources cited |
get_run_summary |
A run’s headline numbers, or the latest completed run of a site |
search_questions |
Search the questions the latest full run asked, with whether the site’s pages answer them |
get_crawler_traffic |
AI crawler and assistant requests the site’s own server saw over 90 days |
get_citations_for_question |
For one question, what each engine answered in recent runs and every source it cited |
Resources: vercatus://sites, vercatus://sites/{siteId} and vercatus://glossary, the terms Vercatus uses
and how to report its numbers.
What the assistant sees
Section titled “What the assistant sees”The tools return the product’s numbers with their denominators and never a secret: the crawler-traffic hook address, integration tokens and webhook secrets stay out. Changes made through the tools are in the audit log as the API key. Answers the assistant writes from these numbers are its own; the Vercatus workspace remains the record.