Proofsource and your own MCP server: your tools, one approved call at a time
Add the address of your own MCP server and Proofsource reads its list of tools. Pick one and you see a dry run that shows exactly what the call would do and sends nothing. Only when a person says Yes does Proofsource send it, once. Your systems join the fix loop without giving an AI free rein over them.
| Tool | What it does | |
|---|---|---|
| update_changelog | Adds an entry to the public changelog | Dry run |
| add_faq_entry | Adds a question and answer to the help center | Dry run |
| list_integrations | Reads the integrations page data | Dry run |
| flag_page_for_review | Marks a page for editors | Dry run |
Proofsource and your own MCP server, in short
The facts, ready to quote.
- Stage
- Live for every workspace, in the Your own tools tab
- Connects with
- The https address of your MCP server, plus a bearer token if it needs one
- Access it needs
- Only what your server lets that token do; Proofsource reads the tool list first and calls nothing yet
- Where it shows
- Integrations, Your own tools tab, and the AI apps and assistants group in the directory
- Safety
- Every call has a dry run and a Yes, and each Yes sends one call
- Disconnect
- Remove the server in the Your own tools tab and Proofsource cannot call it again
Why we love your own MCP server
In the makers' own words, from their own pages. Proofsource builds on it, never around it.
Tools that describe themselves
An MCP server lists each tool with a name, a plain description and the inputs it takes, which is what lets Proofsource show you a dry run before anything is sent.
From modelcontextprotocol.ioOfficial kits for building a server
The Model Context Protocol project publishes software development kits for several languages, so your team can expose its own tools without starting from scratch.
From modelcontextprotocol.ioLocal or remote, your choice
An MCP server can run beside your own systems or be hosted remotely, with standard ways to authenticate, so the tools stay where your data already lives.
From modelcontextprotocol.ioHow Proofsource uses your own MCP server
Lets Proofsource call tools on your own MCP server, with a dry run first and a person's Yes before anything is sent.
- Connect
Add your server
You give Proofsource a name, the https address and, if your server needs one, a bearer token. Proofsource reads the list of tools and nothing else.
- Read
Pick a tool and see the dry run
Choose a tool and fill in its inputs. The dry run states what the call would do and sends nothing to your server.
- Ship
A person says Yes
Only after a person approves does Proofsource send the call, once. The result is saved beside the fix that led to it.
- Prove
See whether it moved the answers
Proofsource asks the same buyer questions again and shows whether ChatGPT, Perplexity, Claude and the other engines now name your brand.
Set up your own MCP server in 4 steps
Open Integrations, Your own tools
The tab is open to every workspace and sits beside the other Integrations tabs.
Add the server
Enter a name, the https address of your MCP server and an optional bearer token. Addresses that are not https are refused.
Let Proofsource read the tools
Proofsource lists what the server offers. If it cannot reach the address, it says so in one plain line and nothing is saved as working.
Run a dry run, then decide
Pick a tool, check the dry run, and approve only the calls you want. You can refresh the tool list or remove the server at any time.
What you unlock with your own MCP server
Your systems in the loop
Reach the tools only you have, such as your changelog, help center or product data, from the same place you approve fixes.
A dry run every time
See exactly what a call would do before it does anything.
Sent once, on a Yes
A person's approval sends one call, never a standing permission.
Your token, your limits
Proofsource can do only what your server lets the token do.
Refresh the tool list
When you add a tool to your server, refresh and it appears.
Built for agents
Pair your tools with agents that draft and propose, while people stay in charge.
Google AI OverviewsGoogle AI Mode
Questions about Proofsource and your own MCP server
What does Proofsource do with my MCP server?
It reads the server's list of tools. When you pick a tool, it shows a dry run of the call and sends nothing. Only after a person says Yes does it send that one call to your server.
What access does it need?
The https address of your server and, if your server asks for one, a bearer token. Proofsource can do only what your server allows that token to do, so give it a token with the narrowest rights that work.
Can a call go out without a person approving it?
No. Every call starts with a dry run, and each Yes sends the call once. A Yes is not a standing permission for later calls.
How do I disconnect it?
Remove the server in the Your own tools tab. Proofsource stops reading its tools and cannot call it again until you add it back.
What does Live mean here?
The Your own tools tab is open to every workspace. The Model Context Protocol itself is an open standard, so any server that follows it can be added.
How does this help my AI visibility?
Fixes often need a change in your own systems, such as a help article or a changelog entry that engines can cite. Calling your own tool with approval gets those changes made, and Proofsource then checks whether ChatGPT, Perplexity, Claude and the other engines start naming you.
What if Proofsource cannot reach my server?
It tells you in plain words that it could not reach a server at that address and asks you to check the address and try again. Nothing is called.
Pairs well with your own MCP server
Put your own MCP server to work for your AI visibility.
Scan your brand, connect your own MCP server, and see what AI says about you on up to 10 engines.
Google AI OverviewsGoogle AI Mode