Proofsource and Claude Code: your AI visibility, in the terminal
Add Proofsource to Claude Code and your terminal can answer what AI says about your brand. Read visibility, competitors and cited sources, turn open gaps into a prioritized fix plan, and mark each fix shipped so Proofsource can check it moved the answers in ChatGPT, Perplexity, Claude and the other engines.
$ claude mcp add --transport http proofsource https://mcp.proofsource.co/mcp Added HTTP MCP server proofsource > /mcp proofsource needs sign-in [Authenticate] proofsource connected
Proofsource and Claude Code, in short
The facts, ready to quote.
- Stage
- Live, with the plugin and the MCP server
- Connects with
- The Proofsource plugin, or one MCP command and a browser sign-in, or an API key
- Skills added
- ai-visibility-audit and fix-plan
- Access it needs
- Read by default; writes ask first
- Where it shows
- Settings, Connect AI apps, For developers
- Disconnect
- Revoke the app under Connected apps, or delete the API key
Why we love Claude Code
In the makers' own words, from their own pages. Proofsource builds on it, never around it.
An agent that lives in your terminal
Claude Code works in the terminal, your IDE, Slack and the web, reading your codebase and carrying out multi-step work with you.
From claude.comConnects to your tools through MCP
One command adds a remote MCP server over HTTP, so Claude Code can reach outside systems and data with a short sign-in.
From code.claude.comPlugins that bundle a whole workflow
A plugin packages skills, agents, hooks and MCP servers into one install from a marketplace, so a whole setup arrives with a single command.
From code.claude.comHow Proofsource uses Claude Code
Read gaps, plan fixes and mark them shipped from your terminal inside Claude Code.
- Know
Ask what AI says about you
The ai-visibility-audit skill returns a sourced summary of your visibility, competitors, cited sources and recent changes, straight from your Proofsource workspace.
- Act
Turn gaps into a plan
The fix-plan skill reads your open gaps and writes a prioritized plan, so the next engineering or content task is obvious.
- Ship
Mark the fix shipped
When you merge the change, Claude Code dates the fix in Proofsource, after asking you first, so the effect can be verified.
- Prove
Check the before and after
Ask Claude Code whether a shipped fix worked and it reads Proofsource's verification, with sample sizes, in the same session.
Set up Claude Code in 4 steps
Install the plugin
Run /plugin marketplace add proofsource/plugins, then /plugin install proofsource@plugins. The plugin adds the MCP server and both skills.
Or add the server directly
Run claude mcp add --transport http proofsource with your server URL from Settings, Connect AI apps, For developers.
Sign in
Run /mcp, pick proofsource and choose Authenticate. A browser opens and you sign in to Proofsource. No API key is needed.
Or use an API key for unattended agents
Create a key under Settings, API Keys and pass it as an Authorization header. Use a write key only where an agent should mark fixes shipped.
claude mcp add --transport http proofsource https://mcp.proofsource.co/mcp
- get_visibility: how often AI answers mention you
- list_fix_opportunities: your gap register
- mark_fix_shipped: date a fix so it can be verified
See it at work
Drawn screens with sample data for an example brand.
What you unlock with Claude Code
Visibility in the terminal
Ask about mentions, competitors and cited sources without leaving your editor flow.
Fix plans from gaps
A prioritized plan from your open gaps, ready to turn into tickets or commits.
Shipped means measured
Mark a fix shipped and Proofsource verifies the before and after.
The MCP server
The same tools every MCP client can use.
Writes ask first
Reading is open; tracking a question or marking a fix shipped always asks.
Sign in or key
Browser sign-in for people, an API key for agents with no one to sign in.
Google AI OverviewsGoogle AI Mode
Questions about Proofsource and Claude Code
What does Proofsource add to Claude Code?
Ten tools, list_brands, list_topics, get_visibility, get_competitors, get_citations, get_changes, list_fix_opportunities, get_fix_verification, create_prompt and mark_fix_shipped, plus two skills called ai-visibility-audit and fix-plan. Together they read your AI visibility, plan fixes and track what shipped.
How do I connect it?
Install the plugin from the marketplace, or add the server with one claude mcp add command, then run /mcp, pick proofsource and choose Authenticate to sign in with your Proofsource account.
Can Claude Code change anything without asking?
No. Reading is open, but tracking a new buyer question or marking a fix shipped always asks you first, and a read-only key cannot write at all.
How do I disconnect it?
Revoke it under Connected apps in Settings and its next request is refused, or delete the API key. Removing the server from Claude Code also stops it.
What does Live mean here?
The MCP connection and the Claude plugin are generally available. The plugin is installed from the Proofsource marketplace listing.
How does this improve my AI visibility?
It shortens the path from finding a gap to shipping the fix and proving it worked. Developers see the same sourced gaps the marketing team sees, in the tool where the code lives.
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 Claude Code
Put Claude Code to work for your AI visibility.
Scan your brand, connect Claude Code, and see what AI says about you on up to 10 engines.
Google AI OverviewsGoogle AI Mode