Snitcher shows you which companies are visiting your website. The new MCP server lets you ask about them from the AI assistant you already work in.
Why we built an MCP server
Snitcher has always answered one question well: which companies are on your site and what they did there, but the answer has always lived in one place.
The place where questions get asked has moved.
Research starts in Claude now, or any other AI assistant. More of the surrounding work (account plans, outbound drafts, pipeline reviews) happens in a chat thread than in any one tool, and through all of it your visitor data has stayed a tab-switch and an export away.
We went in with three beliefs about how that should work:
- Simple: You should be able to ask a question about your data, without needing filters and date pickers before it can be answered
- Complete: The assistant should see what you see (companies, sessions, contacts, segments, CRM state)
- Careful: Anything that writes, spends, or deletes should ask you first
So we built one.
(MCP is the open standard AI assistants use to work with other software. If yours speaks it, it now speaks Snitcher.)
Getting an answer without operating a dashboard
“Which high-intent companies matched our ICP this week?” used to be a workflow rather than a question. Open Snitcher, build the segment, set the date range, scan the list, send it to wherever the actual work was happening.
Now it’s a simple sentence. So is the follow-up, and the one after that:
- “Show me identified companies from the last two weeks that fit our ICP, highest engagement first.”
- “Which of these does our CRM not know about yet?”
- “What did the top three look at while they were here?”
- “Tag them for outreach and push them to the CRM.”
That’s one thread, not four dashboard trips. Behind each one the assistant is running Snitcher’s filters, lookups and integration actions and what comes back is ready for the next question you need answered.
Going from a company to a conversation
A promising account used to kick off a chase. Find the right person, verify an email, piece together which pages made them interesting and then send to your outreach tool of choice.
Now that whole run happens in one thread. The assistant pulls the contacts behind an account, reveals a verified email (it quotes the credit cost first), summarises the sessions that made the company worth a look and drafts the outreach from what they actually read.
Letting an assistant near your pipeline
Giving an assistant write access to your pipeline is the part everyone hesitates on, and fair enough. An API key doesn’t ask follow-up questions, and you find out what an integration changed after it’s changed it.
The server asks instead. It checks with you before changing anything, tells you the price before spending a credit, and runs anything you might regret as a dry run first so you can see exactly what would change before it does.
Wherever the question comes up
Work moves between a dashboard, a CRM, a chatbox, and more every month. We’d rather bring Snitcher along than make you come back to the tab, so “who’s on our site, and are they worth it?” gets its answer wherever it was asked.
The MCP server is the latest step. The same companies that fill the Inbox and sync to your CRM now answer questions from inside your assistant too. Connecting is a one-time job inside your assistant, and the docs walk you through it.
Available now. All it takes is an active Snitcher account and an assistant that supports MCP. Switch it on under Settings → Integrations.
Connect your assistant, start asking →
(The server ships with a feedback tool wired straight to us. If it can’t answer something you wish it could, tell it so, literally.)
(The whole thing started as a proof of concept built on a whim. Andreas, the staff engineer behind it, tells that story in Source Code: Behind the Snitcher MCP.)
