Use Viewfy from Claude,
Cursor or ChatGPT.
Viewfy is an MCP server. Paste one URL into the chat or editor you already have open, sign in when it asks, and he is there: the same reply queue, audits and blog. Nothing publishes without you.
https://api.viewfy.ai/mcpHow do you set it up?
Paste the URL above into your client as a custom connector and sign in with Google when the browser opens. That is the whole setup: no key to create, nothing to configure beyond the URL. The exact steps for each client:
Cursor installs in one click. Dedicated setup lives on /claude and /chatgpt. Those buttons copy the URL and open the page where it gets pasted.
Claude
web and desktop, paid plans- Settings, then Connectors, then Add custom connector.
- Name it Viewfy, paste the URL above, and click Add.
- Click Connect and sign in with Google in the window that opens.
Claude Code
One command in your terminal:
claude mcp add --transport http viewfy https://api.viewfy.ai/mcp
Then run /mcp inside a session to sign in.
Cursor
Add to Cursor in one click, or add this to .cursor/mcp.json in your project, or ~/.cursor/mcp.json for every project:
{
"mcpServers": {
"viewfy": { "url": "https://api.viewfy.ai/mcp" }
}
}Cursor opens the browser to sign in the first time a tool runs.
ChatGPT
developer mode, paid plans- Settings, then Apps, then Advanced, then turn on Developer mode.
- Create a new app: name Viewfy, paste the URL above.
- Complete the sign in, Google preferred.
What can you do from chat?
Everything the web app can do with a product, from the chat you already have open: read his queue, approve or revise drafts, run audits, write posts, and pull yesterday's numbers. The four that matter most:
Approve his replies.
He finds live threads where your buyers are already asking and drafts the answer. From chat you read the queue, revise a draft in plain language, and approve or reject. That approval stays yours unless you hand it over by name.
Read the evidence.
Run the audit on any URL, walk the issues by severity, and see who ChatGPT, Claude and Perplexity name when someone asks the buying questions in your category.
Ship the artifacts.
Write and revise blog posts, preview an SEO fix PR before it opens against your repo, and go through the link outreach queue email by email.
Ask for the morning report.
Yesterday's visitors, signups, citations and what waits on you. The same facts the daily digest sends, available whenever you ask.
Where is the API key?
There is not one. Auth is OAuth 2.1 with PKCE. Your client registers itself, you sign in once in the browser, and it gets a short-lived token with only the scopes you approved. Keys that live in config files leak and never expire; these expire within the hour and rotate on their own.
Scopes are asked for by name: read, write, and policy. Policy is the one that lets an agent publish without a human. It is never granted by default, it never rides along with a blanket grant, and every change made under it writes an audit row you can read.
Never used Viewfy? The sign in creates your account. Connect from your editor, paste your product URL, and start from there.
Questions.
Which clients work?
Anything that speaks Streamable HTTP MCP with OAuth: Claude on web and desktop, Claude Code, Cursor, ChatGPT in developer mode, and most agent frameworks. Plain scripts can use the REST twin at /api/agent instead.
Can the agent post without me?
Not unless you grant the policy scope by name on the consent screen. Out of the box every reply and post waits in the approval queue, and every change to that setting writes its own audit trail.
Is it free to connect?
Connecting and reading are free. Some writes, like publishing blog posts and opening SEO fix PRs, need a paid plan. The retrievability check at viewfy.ai/check-seo-geo stays free.
He needs a product to work on.
Paste your URL and he reads the site, builds the brain, and starts finding the threads where your buyers are asking. You approve, he posts, you ship.