How to Map Warm-Intro Paths to a Target Company in Claude

Before you reach out to a target company's CEO, you want to know whether anyone at the fund already has a warm path in. At most firms the answer takes a one-time export of everyone's connections, or a partner asking around on Slack and hoping someone remembers a relationship from years back. By the time you have an answer the moment to act has often passed.

This guide shows how a PE or VC associate maps warm-intro paths to a target company by chatting with Claude over the Crustdata MCP. You will see the prompts to type, the data that actually surfaces a path, and how to read what comes back, without building a tool or moving the fund's CRM and connection records out of the systems you already own. We are writing this from how a lower-middle-market PE firm we spoke with described the work, and the data method that makes it repeatable. You can try the same thing on a free Crustdata account, with 100 credits to get started at crustdata.com.

What does mapping a warm-intro path to a company mean?

A warm-intro path is a person you can reach who can in turn reach into the target company. Mapping one means finding that person before you send anything cold. For a deal team the target is usually a founder or a CEO, and the path is whoever at the fund, or in the fund's extended network, already knows them or worked alongside them.

The reason this matters is the conversion gap. Warm introductions from people an investor already trusts convert at far higher rates than cold email, which Flowlie puts at 10 to 15 times the response rate against the 2 to 4 percent a cold email tends to get. Commsor reports cold-email response rates below 2 percent, and a mutual introduction lifting that several times over while the deal moves faster. When a connector vouches for you, the recipient extends trust before the first call. The hard part is finding the path quickly enough to use it.

What makes a connection actually warm?

Not every connection is one you can ask. A name in a contact list carries far less weight than a former colleague who would take the call. A warm connection rests on real, checkable signals, and a few of them tend to hold up.

• Overlapping employers and the years: two people who worked at the same company over the same window, in broadly similar roles, very probably know each other. The years are what make it real, since sharing an employer a decade apart means nothing.

• A shared school: classmates and program alumni carry a softer but genuine tie, especially in a tight founder community.

• Recent social engagement: if one person keeps commenting on the other's public posts, the relationship is live rather than years out of date.

These signals are what the data can show you. What the data cannot see is the part that lives only in your fund's own records, the email threads and the meetings that say how warm the relationship really is today. That side stays inside your CRM, and it is yours to read against what the public signals surface. For the full method of scoring each signal and ranking the result, see our guide on mapping who-knows-who.

How do you map a warm-intro path in Claude?

This is the part the firm we spoke with cared about most. They had run a do-it-yourself setup for years and were moving the front end to Claude, while keeping the rest of the stack in place. The associate's question was practical. What do I actually click, and what does my one technical person set up once?

The one-time setup, what you click versus what gets configured

A Claude Code agent with Crustdata's MCP server configured can pull people data for you inside a normal chat, so the associate types questions and Claude runs the lookups. The setup behind that happens once. Your technical person installs the Crustdata MCP server and points your corporate Claude account at it, which is a short configuration rather than a build. After that the associate never touches it again. If you want the step-by-step on the connection itself, we wrote it up in connecting the Crustdata MCP server to Claude.

The point the firm kept returning to is that this does not require an engineering team. A non-technical associate runs the searches, and a single technical helper does the one-time wiring.

The prompt to type

Once the MCP is connected, you ask in plain language. A working first prompt looks like this:

"Find people who now work somewhere else but used to work at [target company]. I want to know who in our network might already know someone there."

Claude reads that, calls Crustdata through the MCP, and comes back with people whose work history includes the target. You can narrow it the same way you would talk to a colleague, by adding a city, a seniority level, or a time window.

Reading the output, and the trust objection

The associate at the firm was openly skeptical that this could work. In his words, "I'll be amazed if the agents can figure that out," because the useful path is often a second-degree connection that nobody on the team even remembers. That doubt is fair, so the answer is to show the surfaced people rather than claim anything magic.

A believable result reads like this. You ask about a target company, and Claude returns a person who is now a VP somewhere else but spent 2014 to 2017 at the target, in the same group as a partner at your fund over those same years. That overlap is the path, and it rests on two work histories that line up on the same employer and the same window, so the match sits in the data where you can check it. You still apply judgment about how warm it really is, but the candidate is in front of you in the time it takes to read a chat reply, instead of after a day of polling colleagues over Slack and hoping someone remembers.

Behind the chat, Claude is running a person search filtered on past and present employers. If you ever want to see the exact call, it looks like this.

The experience.employment_details.company_name field covers every employer a person has had, present or past. So a profile whose present role is elsewhere, returned by this filter, is someone who used to work at the target. Those are your candidate introducers. The chat path runs this for you, and the code is shown only so you can see there is no black box behind the answer.

What does not happen to your CRM

Nothing here writes back to your system of record. The firm we spoke with has kept the same CRM for fifteen years and had no intention of moving off it, and they should not have to. Crustdata supplies the employment and education data that surfaces the overlap, and your relationship records stay exactly where they are. As the firm put it, importing a one-time list of connections gives you a picture that "is not living." The data pull is live, but it adds to your view rather than replacing the records you already trust.

Keeping the path list fresh

A list of who knows whom goes out of date the moment someone changes jobs. Rather than re-running the export every quarter, you can watch for the moves that matter and let the map update itself. Crustdata's Watcher tracks profile and employment changes as they happen, so a new overlap appears when a person joins or leaves a company rather than waiting for your next manual refresh. The teams we spoke with felt the manual version most when a path went out of date right when they needed it. Try the first search free with 100 credits.

What if you want to script it instead of chatting?

If your one technical person would rather run this in code, the same lookup is a direct API call. You filter people search on past employer to get your candidate introducers, then enrich each one for the full work history and education that confirm the overlap.

From there you pass each profile URL to people enrichment for the dated employment and education detail, and apply your own rule for what counts as a warm-enough overlap. You can weigh recent social engagement into that rule as another signal of a live relationship. The chat path and the code path hit the same data, so the only real choice is how much you want to run yourself.

Where to start this week

The fastest way to see whether this works for your fund is to run it on one target. Pick a company you are about to approach, ask Claude who in your network has worked there, and read the overlap it returns against what you already know about your partners' backgrounds. That single search tells you more than a week of asking around, and it costs you nothing to try.

From there, the build-versus-buy question is smaller than it looks. A product lead and a single engineer can wire the API path in an afternoon, because the hard part was never the code. It was getting employment data fresh enough and connected enough to trust, and that is the part you can now buy. Crustdata is the data layer for investment teams mapping their own warm-intro paths. So come and see what the overlap looks like across your own target list. Start free with 100 credits at crustdata.com, or book a demo to walk through the setup with us.

Frequently asked questions

Does this move my CRM out of the systems I already use? No. Your relationship records and your system of record stay exactly where they are. Crustdata supplies the employment and education data that surfaces a path, and nothing here writes back to your CRM. You read the public overlap against your own records, side by side.

Can it really find second-degree connections nobody remembers? It surfaces people in your network who have worked at, or alongside people at, the target company, which is where most forgotten paths hide. A true third-degree link that exists in no data still depends on the connection records you feed in. The honest version is that it finds the checkable overlaps fast and leaves the unrecorded ones to you.

Do I need an engineer to do this? Not for the chat path. A non-technical associate runs the searches once the Crustdata MCP server is connected to your Claude account, which is a one-time setup your technical person handles. You only need to write code if you choose the direct API route instead.

What data is it using to find the path? Entity-resolved employment history with dates, education, and public social activity. Your private email and CRM stay untouched unless you choose to connect them. The warm-path signal comes from where people have worked and studied, matched against your own team's backgrounds.

How do I keep the connection list from going out of date? Watch for the job moves that change the map instead of re-exporting it. Crustdata's Watcher tracks employment and profile changes as they happen, so a new overlap shows up when someone joins or leaves a company, rather than waiting for your next manual refresh.