Prospecting with Claude: a tutorial to find, write to and reply to prospects through MCP

Prospecting with Claude means asking your AI assistant, in plain English, to do what you would do by hand in a prospecting tool: “find me owners of small manufacturers hiring a maintenance manager”, “rewrite this message, shorter”, “who replied yesterday?”. Claude does not find prospects on its own. Connected through an MCP connector to a business signals tool, it searches, writes, replies and learns from your feedback. Here is the full tutorial, the use cases and the limits to know.

📌 Key takeaways

• Claude has no prospects: it needs a connector to a tool that detects business signals on LinkedIn and elsewhere.
• Connecting takes five minutes: one address to paste into Claude, one sign-in in your browser, and you are ready.
• Six everyday uses: search for prospects by signal, write the messages, reply to conversations, sharpen the targeting, add prospects from elsewhere, track what works.
• Claude acts with your rights: at first, you approve everything before it is sent. Access is revoked in one click.

Prospecting with Claude: what does it really change?

Prospecting with Claude replaces the clicks and forms of a prospecting tool with a conversation. You say what you are looking for, Claude picks the actions to run in the tool, shows you the result and asks for your approval before anything reaches a prospect.

The link goes through MCP (Model Context Protocol). The official site describes it as an open-source standard for connecting AI applications to external systems, and compares it to a USB-C port. A Claude connector is that plug: the tool exposes a list of actions (“search for prospects”, “draft a reply”…), and Claude uses them to answer your request.

Sales teams are moving fast. According to Salesforce’s 2026 report, 54% of sellers have already used AI agents and 55% use AI for prospecting; they expect agents to cut prospect research time by 34% and email drafting by 36%. 🚀

But Claude alone does not know who is hiring this week or who just changed jobs. A business signals tool brings the material, Claude brings the conversation. For the method, see how to identify business signals.

Bottom line: Claude is the interface, not the source. The quality of your prospecting with Claude depends on what the connector gives it to read.

How do you connect MeetMagnet to Claude, step by step?

You connect MeetMagnet to Claude by adding a custom connector with the MCP server address, then authorising access in your browser. No key to copy, no code. ⏱️

1. In MeetMagnet, open Settings then AI integrations. The “MCP server” block shows the address to copy: https://app.meet-magnet.com/mcp. Access is limited to your account and can be revoked at any time.

2. In Claude, open Settings then Connectors. Depending on the version, the menu is also called “Customize”. Click Add, then Add custom connector.
3. Name the connector “MeetMagnet” and paste the address. Click Continue.

4. Sign in to your MeetMagnet account. OAuth authentication opens in your browser: you approve the permissions on MeetMagnet’s consent screen. Claude never sees your password.
5. Turn the connector on in the conversation. Anthropic’s help center says to use the + button at the bottom left, then Connectors, to switch it on or off for each conversation. Ask your first question. ✅

Which Claude plans allow it? According to Anthropic’s help center on custom connectors, the Free, Pro, Max, Team and Enterprise plans, on the web, Claude Desktop and Cowork. The free plan is limited to one custom connector. On Team and Enterprise, an owner adds it for the organisation first, then each member connects to it.

In Claude Code, the documentation gives the command to run in the terminal: claude mcp add --transport http meetmagnet https://app.meet-magnet.com/mcp, then you sign in from the /mcp command. The same address works in ChatGPT, Mistral or Perplexity: the exact path for each assistant is on our MCP connector page. 🔌

What can you ask Claude to do for prospecting?

You can ask Claude to do anything the MeetMagnet app does. Here are six uses, each with a typical request and what happens behind it.

1. Find prospects by business signals

Describe your customer and the moment they need you. Claude reads your setup (target, offer, tracked intents), then suggests new intents: an interest (“interested in predictive maintenance”) or a moment (“hiring a maintenance manager”), with the matching LinkedIn searches.

Example

You: “I sell maintenance software. Find me owners of small manufacturers who are hiring a maintenance manager or talking about machine breakdowns.”

What happens: Claude creates a “moment” intent on the hiring and an “interest” intent on machine downtime, adds the LinkedIn searches, and checks that the target (job titles, industries, size, region) matches your request. Prospects then land in “To review”, each with their signal and a score.

Searches are run by our robots, not by your LinkedIn account: this is not a “LinkedIn MCP” crawling the network with your login. An account can track up to 15 intents and 45 active searches. Other sources (job ads, reviews, press) come with the multi-source option of the Managed plan. 🎯

2. Send highly personalised messages that do not sound like AI

For each prospect, MeetMagnet prepares a sequence on LinkedIn, by email or both. From Claude, you review it, rewrite it and approve it, and you set the channels, the delays and the tone of your openers.

Example

You: “Show me the prospects to review. For each one, rewrite the first message: shorter, without quoting their post word for word.”

What happens: Claude lists the pending prospects with their signal, rewrites the drafts and shows them to you. It only approves them once you agree, and only messages that have not gone out yet can be edited.

Watch out for the hyper-personalisation trap: a message that piles up the post, the career path and the year the company was founded smells like a machine. Our advice fits in four points: one useful piece of information, described the way you would say it out loud, your own tone, and sometimes no signal mentioned at all. It is all in our article on hyper-personalised outreach that sounds like AI. ✍️

Example

Signal: the owner of a small engineering firm posts a job ad for a maintenance manager.

Opener that misses: “Hi Mark, I saw your maintenance manager job ad posted on Tuesday. Your 80-person company, founded in 1987 and specialised in aluminium machining, seems to be growing fast.”

Opener that gets a reply: “Hi Mark, when a plant hires in maintenance, it is often because machine downtime is getting expensive right now. Is that your case, or is it mostly someone leaving?”

The first one recites the profile; the second talks about their problem and asks a simple question. 💡

3. Reply to prospects and manage conversations

When a prospect replies, MeetMagnet classifies the reply and calculates a score from 0 to 100 based on their profile, their interest and the conversation. From Claude, you handle the warmest ones first.

Example

You: “Who replied to me this week? Rank them from warmest to coldest and suggest a reply for each.”

What happens: Claude rereads each conversation, drafts a suitable reply (two time slots for a yes, a question for someone hesitating) and shows it to you. It only sends it once you agree, on LinkedIn or by email depending on the best channel.

Claude can also attach a PDF brochure, an image or a video (three files at most), and rewrite or cancel a follow-up that is already scheduled. ⚠️ One detail to know: a message sent by hand cancels the automatic follow-ups planned for that prospect, and Claude has to tell you before sending. To go further on handling these exchanges, see our AI conversational agent.

4. Let the account learn: targeting that gets sharper

This is the most underrated use. Every piece of feedback you give Claude becomes a rule for the account: a prospect excluded with a reason (wrong job title, competitor, existing customer, size, industry, location), an intent that brings noise, a misclassified reply. At first, every approval trains the AI; then the target sharpens on its own.

Example

You: “These three prospects are not a fit: two are consultants, the third is a competitor. Exclude them and tell me which intent brought them in.”

What happens: Claude excludes them with the right reason, which cancels their drafts, finds the original intent and suggests adjusting it rather than turning it off if it already brings replies.

Intents that bring results are protected: Claude does not turn off a “high-performing” intent without your explicit agreement. It can also classify replies that were never analysed (no, price request, time slot). 🔁

5. Add prospects from elsewhere

Someone you met at a trade show, an owner quoted in the press, a list of former customers: Claude adds them to MeetMagnet from their LinkedIn profile, with the context that explains why to contact them. The record lands in “To review” and the sequence is prepared like the others.

Example

You: “Add this owner I met at the trade show: linkedin.com/in/… He told me about a new warehouse planned for March. Here is his email too.”

What happens: Claude creates the prospect with that signal (a source outside LinkedIn), their email for the email steps, and starts writing the sequence.

For a list, Claude adds the profiles one by one. It can also correct a record (job title, company, signal) or archive a contact. Influencers whose audience interests you are added in the app, in the Targeting menu. 📥

6. Track what works

Ask Claude what worked best this month: it suggests changes to the target or the messages, which you approve one by one. If your CRM also has a connector, Claude adds the prospects who said yes, as in our article on the small business CRM connected to Claude. 📊

What you ask Claude, what MeetMagnet does

What you ask	What MeetMagnet does
“Find HR directors hiring in Manchester”	Creates the intent and the searches, then brings back prospects with their signal
“Show me my current target”	Shows job titles, industries, sizes, locations, intents and active searches
“Rewrite this message, shorter”	Edits the draft or the follow-up not yet sent
“Approve the sequences for these prospects”	Schedules sending, within the account’s LinkedIn limits
“Who replied? Suggest a reply”	Ranks replies by score and drafts an answer for each
“Send the reply to Claire”	Sends on LinkedIn or by email, once you agree
“This prospect is a competitor, exclude them”	Excludes them with the reason and refines the targeting
“Add this profile I met at the trade show”	Creates the record with its signal and prepares the sequence

Prospecting with Claude: what are the limits and risks?

The first limit is simple: Claude acts with your rights. What it does in MeetMagnet, you did. Hence three rules.

• ✅ Approve before sending, at first. Claude drafts, you review. Once the messages suit you, automatic mode is set in the app’s settings.
• ✅ Only plug in connectors you trust. Anthropic’s help center points out that a custom connector is not verified by Anthropic, and advises reviewing the requested permissions and only allowing “always” for tools you trust.
• ❌ Avoid tools that drive LinkedIn with your login. This is the most common mistake: a “LinkedIn MCP” server found online, which browses with your session, can trigger a restriction. Our article on the restricted LinkedIn account covers the causes and the limits. With us, searches do not go through your account, and sending stays under 70 invitations a week, spread between 8 am and 6 pm on weekdays.

Then there is the data question. What Claude reads to answer you goes through Anthropic, a US company; if that is a blocker, the same connector works in Mistral’s Le Chat. 🔒 And access can be revoked at any time: by disconnecting the connector in Claude, or from MeetMagnet.

One last limit: with a fuzzy target, Claude just produces fuzz faster. A well-tuned AI SDR always starts with a precise target.

How much does prospecting with Claude cost?

Two subscriptions to plan for: Claude’s, and that of the tool plugged into it.

• Claude: the free plan is enough for one connector; paid plans accept more.
• MeetMagnet: searching for prospects from your AI is free to start. To prospect and track your results, you need an account: Self-serve at €149 excl. VAT per month, with a free trial, or Managed at €299 excl. VAT per month, where someone on our team calibrates the target and messages with you. Details are on our pricing page.
• To connect the whole team to Claude, the CRM, the calendar and MeetMagnet, our AI integration is quoted on request.

For other uses of AI in sales, see our guide to AI prospecting. 👋

Prospecting with Claude replaces neither your judgement nor your offer: it removes the clicks between the idea and the action. Plug in a connector that brings fresh business signals, start by approving everything, give feedback on every prospect you discard, and let the account learn. After a few weeks, you are no longer talking to a tool, but to someone who knows your market.

Sources

• “What is the Model Context Protocol (MCP)?”, Model Context Protocol, accessed 6 November 2026, https://modelcontextprotocol.io/docs/getting-started/intro
• “Get started with custom connectors using remote MCP”, Claude Help Center, accessed 6 November 2026, https://support.claude.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp
• “Connect Claude Code to tools via MCP”, Claude Code Docs, accessed 6 November 2026, https://code.claude.com/docs/en/mcp
• “The Productivity Gap: New Survey Shows 9 in 10 Sellers Are Betting on AI and Agents To Help”, Salesforce, 3 February 2026 (survey of 4,050 sales professionals, August to September 2025), https://www.salesforce.com/news/stories/state-of-sales-report-announcement-2026/
• “MCP connector” page, MeetMagnet, accessed 6 November 2026, https://www.meet-magnet.com/en/product/mcp/