An MCP server for your AI visibility data
The Proofsource MCP server lets any MCP client, such as ChatGPT, Claude, Claude Code, Cursor or an agent you build, read how up to 10 AI engines answer your buyer questions. Ask in plain words and the assistant returns your mention rate, the competitors named instead of you, the sources cited and what changed, each with its sample size.
$ claude mcp add --transport http proofsource https://mcp.proofsource.co/mcp Added HTTP MCP server proofsource > /mcp proofsource needs sign-in [Authenticate] proofsource connected (19 tools)
AI visibility MCP server, in short
The facts, ready to quote.
- Server
- One remote MCP server over Streamable HTTP at mcp.proofsource.co/mcp
- Sign-in
- OAuth sign-in for ChatGPT, Claude, Claude Code and Cursor, or an API key as a bearer token
- Read tools
- Brands, topics, visibility, competitors, citations, changes, scan status, report link and shopping reports
- Engines measured
- Up to 10 AI engines, depending on your plan
- Cost of a read
- Read tools serve answers already collected and start no new scan
- Scope
- A connection reads one workspace, with the role of the person who connected it
- Rate limit
- 60 tool calls a minute per key
- Disconnect
- Revoke any connected app in Settings, AI agents, and its next call is refused
What is an AI visibility MCP server?
The Model Context Protocol (MCP) is an open standard that lets an AI assistant call tools on another service. An MCP server publishes a list of tools with names, inputs and outputs, and any MCP client, such as ChatGPT, Claude, Claude Code or Cursor, can call them in the middle of a conversation. The assistant decides which tool fits the question, so the person never has to name one.
An AI visibility MCP server publishes tools about how AI answer engines talk about a brand: how often the brand is mentioned, which competitors are recommended instead, which pages are cited as sources, and what moved since last week. It turns a dashboard into something you can ask from wherever you already work, in a chat, in an IDE or inside an automated agent.
Proofsource runs that server for every workspace. Each tool reads the same measured answers you see in the app, from every engine your plan covers, and returns percentages with the number of answers behind them, so an assistant can tell a solid trend from a small sample.
How it works
One hosted MCP server that lets ChatGPT, Claude, Cursor or your own agent read your AI visibility, competitors, citations and changes.
Add the server URL
Add mcp.proofsource.co/mcp as a remote MCP server or custom connector in ChatGPT, Claude, Claude Code, Cursor or any MCP client.
Sign in or paste a key
Clients that support MCP sign-in open a browser window: sign in with your email code and click Allow. Others send a read or write API key from Settings.
Ask in plain words
Ask how often ChatGPT names you, who Perplexity recommends instead, or which sources Claude cites. The assistant picks the tool.
Get numbers with their samples
Every rate comes back with the answers behind it, broken down by engine, question or location, plus a visibility card where the client can show one.
Revoke any time
Settings, then AI agents, lists every connected app with who connected it and when it was last used. Revoke cuts it off on the next call.
See it in the product
Screens from the Proofsource app, drawn with sample data for an example brand.
| Rank | Cited source | Answers citing | Type |
|---|---|---|---|
| 1 | zapier.com/blog/best-ai-meeting-assistant | 88 | Listicle |
| 2 | reddit.com/r/productivity | 61 | Forum |
| 3 | otter.ai/pricing | 47 | Competitor |
| 4 | g2.com/categories/ai-meeting-assistant | 39 | Review site |
| 9 | driffle.ai/features | 14 | Your site |
| App | Connected by | Last used | |
|---|---|---|---|
| ChatGPT | maya@driffle.ai | Today, 9:12 AM | Revoke |
| Claude | sam@driffle.ai | Yesterday | Revoke |
| Cursor | dev@driffle.ai | 3 days ago | Revoke |
 Claude Code | dev@driffle.ai | Oct 30, 2026 | Revoke |
Every capability, live today
15 shipped capabilities behind AI visibility MCP server, each in the product now.
Hosted MCP connector
Every workspace gets a ready MCP connector with Proofsource tools, so any MCP client or AI agent can read your AI visibility without custom code.
MCP endpoint over Streamable HTTP
A standard remote MCP endpoint that accepts an API key or OAuth sign-in, applies a rate limit and lists its read tools to any client.
Brand list and search
Lists the brands your workspace tracks so an assistant can pick the right one, and searches them as you type so a chat composer can at-mention the right brand.
Tracked questions
Shows the topics and buyer questions tracked for a brand, so the assistant knows what the numbers are measured on.
Visibility reads
Says whether AI answers mention your brand, by engine or by question, with sample sizes and a visibility card, and a per-location breakdown for brands tracked in more than one place.
Competitor reads
Shows which competitors AI answers recommend instead of you, and how often each one is named.
Citation reads
Ranks the sources AI answers cite for your questions and shows where your own domains rank among them.
Change reads
Reports what moved over a trailing window: your visibility trend, answer churn, competitor changes and new sources.
Scan status
Tells the assistant how far a brand's first scan has got, so it can say when numbers will be ready.
Report link
Returns a shareable link to the brand's report, ready to paste into an email or a chat.
Example report
Shows a worked example report to someone who has not added a brand yet, without signing in.
Shopping brands and merchants
Returns the brands AI shopping answers show for your product questions and the merchants they send buyers to.
Shopping products and keywords
Returns the products AI shopping answers list, so an agent can see where yours appear, with the shopping keyword data behind your product questions.
Agency-scoped connector
For agencies, tool calls are limited to the brands each person may see, and the connector carries the agency's product name.
OAuth sign-in for MCP clients
MCP clients connect to Proofsource through a standard OAuth 2.1 sign-in, so assistants reach your data with your approval.
Who it is for
Ask about AI visibility from the chat you already use
Ask ChatGPT or Claude how your brand did this week and get the mention rate, the competitors and the cited sources without opening another dashboard.
Give your coding assistant real AI search data
In Claude Code or Cursor, check which pages AI answers cite before you rewrite them, with the numbers in the same window as the code.
Answer client questions in seconds
Ask for one client's competitors or sources in chat, and the connector keeps each person inside the brands they are allowed to see.
Check AI shopping answers on demand
Ask which merchants and products AI shopping answers show for your category and where your products appear.
Why it matters
Answers where you work
No context switch: the numbers come into the chat, IDE or agent you already use.
Sample sizes travel with every rate
Each percentage carries the answers behind it, so an assistant does not over-read a small sample.
Reads start no scan
Read tools serve answers already collected, so asking twice costs nothing extra.
Your roles still apply
A connection acts with the connecting person's role in one workspace, and you can revoke it from Settings.
Google AI OverviewsGoogle AI Mode
AI visibility MCP server: Proofsource and other tools
Several AI visibility and SEO platforms now offer an MCP server. Here is what each one lists on its own pages, next to how Proofsource does it.
| Capability | Proofsource | Profound | Peec AI | Otterly.AI | Semrush | Ahrefs | SE Ranking |
|---|---|---|---|---|---|---|---|
| Hosted MCP server | YesOne remote server for every workspace, readable from any MCP client. | YesMCP server gives ChatGPT, Copilot and coding assistants access to Profound dataSource | YesMCP server works with Claude, Cursor, VS Code, Windsurf and n8nSource | YesMCP server for brand reports, prompts, citations, recommendations and auditSource | YesOfficial Semrush MCP for Claude, Claude Code, Cursor, VS Code and ChatGPTSource | YesHosted remote MCP with Brand Radar toolsSource | YesMCP server for rankings, backlinks, audits and AI visibilitySource |
| Sign in without an API key | YesOAuth sign-in with an email code from ChatGPT, Claude, Claude Code and Cursor. | YesOfficial Claude connector, no API keySource | Not listed | YesMCP server with OAuthSource | YesOAuth 2.1Source | Not listed | Not listed |
| Sample sizes, chat cards and shopping reports | YesEvery rate carries its answer count, ChatGPT and Claude can show a visibility card, and shopping data is available as tools. | Not listed | Not listed | Not listed | Not listed | Not listed | Not listed |
| Visibility, competitors and citations as tools | YesTools for visibility, competitors, citations and changes read your numbers | RelatedGives ChatGPT, Copilot and coding assistants access to Profound dataSource | YesWorks with Claude, Cursor, VS Code, Windsurf, n8n; included on paid plans; bulk editsSource | YesBrand reports, prompts, citations, recommendations, audit; OAuthSource | RelatedOfficial Semrush MCP: Claude, Claude Code, Cursor, VS Code, ChatGPT; OAuth 2.1; uses API unitsSource | YesHosted remote MCP lets ChatGPT, Claude or Copilot query Ahrefs; Brand Radar tools includedSource | YesQuery rankings, backlinks, audits and AI visibility from Claude, Codex, Cursor, ChatGPTSource |
| Shareable report link | YesA tool returns a link to the brand report to paste into an email or chat | RelatedDocument management with external sharing to people outside orgSource | Yes80 widgets, shareable with team or public linksSource | RelatedDocument or presentation type, optional own logoSource | YesCreate a shareable online dashboard from a reportSource | Not listed | Not listed |
| Results split by location | YesVisibility adds a per-location breakdown for brands tracked in more than one place | YesMonitoring across 150+ regionsSource | RelatedBrand insights split by countrySource | RelatedFilter by countrySource | RelatedGeographic breakdown of where the brand is mentionedSource | RelatedPrompts run in the location of their source keyword; country list not statedSource | RelatedBy city and engine in SE Visible; also listed as coming soonSource |
Where each of them shines
A broad agent platform with an MCP server, a Claude connector and an agent API for building workflows on Profound data. Compare in full
An MCP server that works across many developer clients, including n8n for automations and bulk edits. Compare in full
An MCP server that also exposes recommendations and audit results, signed in with OAuth. Compare in full
An official MCP server with OAuth 2.1 sign-in that works with Claude, Cursor, VS Code and ChatGPT. Compare in full
A hosted remote MCP server that lets ChatGPT, Claude or Copilot query Brand Radar and other Ahrefs data. Compare in full
One MCP server across rankings, backlinks, audits and AI visibility, alongside a full API.
Questions about AI visibility MCP server
What is the Proofsource MCP server URL?
The server is at https://mcp.proofsource.co/mcp. Add it as a remote MCP server (Streamable HTTP) or a custom connector in your client.
Does it work with ChatGPT, Claude, Claude Code and Cursor?
Yes. ChatGPT, Claude, Claude Code and Cursor connect with a browser sign-in. Any other MCP client works too, either with MCP sign-in or with an API key sent as a bearer token.
Which AI engines does the data cover?
Up to 10 AI engines: ChatGPT, Perplexity, Claude, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, Grok, DeepSeek and Meta AI, depending on your plan. The tools return the answers Proofsource measured on those engines for your tracked buyer questions.
Does asking a question start a new scan or use credits?
No. Read tools serve answers Proofsource already collected for you, so asking about visibility, competitors or citations starts no new scan.
Can the assistant change anything in my workspace?
Read tools only look. A few tools can add a question or date a fix, and those need write access; ChatGPT and Claude ask you before a tool makes a change.
Does Proofsource see my chats?
No. When the assistant calls a tool, Proofsource receives only that call's inputs, such as a brand or a question, and sends back the result.
How is this different from the Semrush or Ahrefs MCP servers?
Those servers are a strong fit if you want AI visibility next to a large SEO dataset. The Proofsource server is focused on measured answers from up to 10 AI engines for your own buyer questions, with sample sizes on every rate and tools that follow fixes through to verification.
Can an agency limit which brands a person can query?
Yes. On an agency workspace, tool calls are limited to the brands each person may see, and the connector shows the agency's product name.
How do I disconnect an assistant?
Open Settings, then AI agents, find the app under Connected apps and click Revoke. Its next call is refused until someone signs in again.
Keep exploring
Try it on your own brand.
Proofsource is live: scan your brand and see AI visibility MCP server at work with your own answers, competitors and sources.
Google AI OverviewsGoogle AI Mode