Proofsource and MCP: one server, every agent
If an agent speaks the Model Context Protocol, it can use Proofsource. One server over Streamable HTTP exposes your brands, topics, visibility, competitors, citations, changes and fix opportunities, plus two careful writes. Sign in with OAuth or use a workspace API key, and every client sees the same sourced numbers.
https://mcp.proofsource.co/mcp
| Tool | What it does | Access |
|---|---|---|
| get_visibility | How often AI answers mention you | Read |
| get_competitors | Share of voice against competitors named | Read |
| list_fix_opportunities | Your gap register and the fix | Read |
| create_prompt | Track a new buyer question | Write |
| mark_fix_shipped | Date a fix so it can be verified | Write |
Proofsource and MCP clients, in short
The facts, ready to quote.
- Stage
- Live
- Connects with
- POST /mcp over Streamable HTTP, with OAuth or a workspace API key
- Tools
- Ten: eight read tools and two writes
- Sign-in
- OAuth with PKCE, client metadata documents and dynamic registration
- Where it shows
- Settings, Connect AI apps, For developers; Settings, API Keys
- What it reads
- Brands, topics, visibility, competitors, citations, changes, fix opportunities and fix verification
Why we love MCP clients
In the makers' own words, from their own pages. Proofsource builds on it, never around it.
An open standard for connecting AI
MCP is an open-source standard for linking AI applications to data, tools and workflows, often described as a USB-C port for AI.
From modelcontextprotocol.ioA clear client and server model
Hosts, clients and servers each have a defined role, with tools, resources and prompts as shared building blocks, so any client can discover what a server offers.
From modelcontextprotocol.ioAuthorization done properly
The specification builds on OAuth 2.1, with discovery, scoped access and audience-bound tokens, so connecting a server stays safe by design.
From modelcontextprotocol.ioHow Proofsource uses MCP clients
Any MCP-capable agent can read your AI visibility and track fixes through one standard server.
- Connect
Point the client at one URL
Use your server URL from Settings, Connect AI apps, For developers. Clients that sign in with OAuth open a browser; unattended ones send a Bearer key.
- Know
Read what AI says
Tools return visibility by engine or question, competitor share of voice, cited sources and what changed, with sample sizes.
- Act
Find and track
list_fix_opportunities returns the gap register, and create_prompt tracks a new buyer question.
- Prove
Date and verify
mark_fix_shipped dates a fix, and get_fix_verification returns whether it worked.
Set up MCP clients in 4 steps
Copy the server URL
Open Settings, Connect AI apps, For developers and copy the Streamable HTTP server URL.
Add it to your client
Add a remote MCP server of type HTTP with that URL. Most clients then open a Proofsource sign-in; no key is needed.
Or create an API key
For agents with no one to sign in, create a key under Settings, API Keys and send it as an Authorization Bearer header.
List the tools
Ask the client to list tools. You should see list_brands, list_topics, get_visibility, get_competitors, get_citations, get_changes, list_fix_opportunities, get_fix_verification, create_prompt and mark_fix_shipped.
| App | Person | Access | Last used |
|---|---|---|---|
 Claude Code | maya@driffle.ai | Read and write | Today |
| Cursor | dev@driffle.ai | Read | Yesterday |
 VS Code | sam@driffle.ai | Read | 2 days ago |
| Nightly report agent | API key | Read | 6 hours ago |
What you unlock with MCP clients
One server
Any MCP client gets the same tools and the same numbers.
OAuth or key
Sign in with a browser, or use a key for unattended agents.
Safe writes
Only two tools write, and a read-only key cannot call them.
Revoke any app
Connected apps lists everyone signed in, with a one-click revoke.
Same data as the API
Build your own tooling on the same sourced numbers.
Ask in plain words
Agents answer questions about your data in plain language.
Google AI OverviewsGoogle AI Mode
Questions about Proofsource and MCP clients
What is an MCP client?
Any app or agent that speaks the Model Context Protocol, such as an AI assistant, an editor or your own program. It connects to servers like Proofsource and calls their tools.
What can an MCP client do with Proofsource?
It can list your brands and topics, read visibility, competitors, citations and changes, list fix opportunities, read fix verification, track a new buyer question and mark a fix shipped.
How does it sign in?
With OAuth, where the client opens a Proofsource sign-in in a browser, or with a workspace API key sent as a Bearer header. Reads need no write permission.
Can an agent change my data?
Only in two ways: tracking a new buyer question and marking a fix shipped. Those need a write permission, and well-behaved clients ask you first.
How do I disconnect a client?
Revoke it under Connected apps, or delete its API key. The next request is refused.
What does Live mean?
The server is open to every workspace today. Clients simply use the protocol version they already speak.
How does this help my AI visibility?
Every agent your team uses can read the same gaps and fixes, so improvements in ChatGPT, Perplexity and the other engines are planned, shipped and verified from wherever people already work.
Which AI engines does Proofsource check?
Up to 10 depending on your plan: ChatGPT, Perplexity, Claude, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, Grok, DeepSeek and Meta AI. The free scan and trial run four.
Pairs well with MCP clients
Put MCP clients to work for your AI visibility.
Scan your brand, connect MCP clients, and see what AI says about you on up to 10 engines.
Google AI OverviewsGoogle AI Mode