12 Claude Prompts for VC Investors and Dealflow Teams
Most dealflow software in 2026 is a CRM with an AI sidebar. Useful, but it doesn't change the actual sourcing motion. The prompts below do. Each one is a real workflow a partner or platform lead at a seed-to-Series-B fund can run today through Claude Desktop with the LinkFetch MCP server connected. Copy, paste, swap the bracketed bits. That's it.
LinkFetch is a compliance-first LinkedIn data API for AI agents, purpose-built for Claude and MCP. The prompts assume you've installed the server once and connected your API key. Every prompt also lists the tools it touches and a credit estimate, so you can budget the workflow before you run it.
Why prompts beat dashboards for VC workflows
VC dealflow software in 2026 has converged on a similar shape: a CRM core, a relationship graph, a few enrichment integrations, and an AI chatbox stapled to the side. 85% of private capital dealmakers now use AI to automate daily tasks, according to Affinity's 2026 toolkit roundup, with 82% of firms running AI on deal sourcing research specifically.
The trouble: those are mostly read paths into a CRM. They surface what's already in your records. The actual sourcing edge in 2026 is in the in-between, the founder who hasn't filed a deck yet, the senior IC at a portco who just rewrote her About to say "exploring next thing", the hiring spike at a series A that telegraphs an imminent round. Those signals don't live in your CRM. They live in LinkedIn, and they don't surface unless you ask.
A prompt running on a cadence asks. A dashboard waits.
How to use this list
Drop each prompt into Claude Desktop with the LinkFetch MCP server connected. The cadenced ones (weekly, daily, quarterly) belong in Claude scheduled tasks, so they run without you. The on-demand ones go into your daily flow as one-off asks. None of them require code, a CRM integration, or a custom build.
A few rules:
- Replace the bracketed bits before you run.
[thesis],[portco],[company]are placeholders; the prompt won't work as written without your inputs. - Watchlist size drives credit cost. A 200-name founder watchlist is cheaper than scanning all of LinkedIn; the prompts assume you've curated.
- LinkFetch is flat per request, no per-result math. The credit estimates are conservative and assume modal coverage on your watchlist.
If you want the assembled routine that chains several of these prompts into a single Monday morning agent, the full Monday dealflow playbook walks through the four-step version most of our seed-fund users run.
What you'll need set up
Three things, all one-time:
- A LinkFetch API key. The free trial gives 500 credits, enough to run most of these prompts at least once.
- The LinkFetch MCP server connected to your Claude Desktop. The setup walkthrough takes under five minutes.
- A watchlist or two. Dealflow watchlists are personal; most partners we work with keep one for thesis companies, one for founder-track ICs, and one for the portfolio.
The 12 prompts
Each prompt below is reproduced in the frontmatter for copy-paste, and explained in plain language here. The credit estimates assume Pro tier list prices and modal coverage on a 50 to 200-row watchlist.
1. Spot a founder before the deck reaches your inbox
A weekly scan of senior ICs whose About copy started reading like pre-launch language. The trigger words ("stealth", "building", "exploring", "next chapter") are deliberately broad. False positives beat false negatives here, you'd rather see one extra name than miss the one that mattered.
The output is a flat list with the exact phrase that fired, so a partner can scan in under three minutes and hand off the genuine leads. Run it Monday morning and the partnership has a shared sourcing layer for the week.
2. Map the alumni network of a portfolio company
Useful before a board meeting, after a layoff round, or when a portco asks the partnership for help replacing a senior leaver. The output shows where alumni land: another portco, a competitor, a brand-new company, or a non-portfolio rocketship.
The "started a company" bucket is where this prompt earns its keep. Founders coming out of a strong portco are pre-qualified by association. Catching them at week one of formation, before they're shopping a deck, is exactly the slot most funds say they want and don't actually staff for.
3. Track a thesis area without reading 800 LinkedIn posts
The hardest part of running a thesis is the read tax. You committed to following 60 companies in some adjacent space, and now you owe yourself 60 LinkedIn scans a week. Nobody does this; it gets dropped by week three.
This prompt does it for you, but it filters: only companies that crossed a real threshold in the last 7 days. Headcount up by 5+, new senior hire, hiring spike. The signal-to-noise is the entire point. Banner changes don't count.
4. Pre-meeting brief on a founder, in 30 seconds
The "I have a meeting in 20 minutes and I have not done my homework" prompt. Returns a complete work history, every prior company started, who worked under them, and any role-change in the last 90 days. End-cap is three meeting questions that aren't on the deck.
The "do not invent biographical detail" line is load-bearing. LLMs left unconstrained will fill gaps with plausible fiction; the prompt forces the model to stay within what LinkFetch returned. If you ask "where did they grow up", the answer should be "not in the data".
5. Headcount-delta heatmap across a competitive set
The competitive intelligence prompt. Once a quarter, run this across the 10 to 15 companies you track in a category and you have a paragraph for the LP letter without spending a partner-hour on it.
Headcount data has been the most reliable B2B funding-stage signal for the last three years, more reliable than press release frequency, website traffic, or job-posts-per-week. The prompt leans into that. The output is a table; LPs read tables faster than prose.
6. People who left a target portco in the last 90 days
The accidental-founder radar. When a senior PM, EM, or designer leaves a hot portco, a third of them will be cofounding something within 12 months. Most won't post about it for six months. The data is in LinkedIn, but you have to ask weekly to catch the window.
The "seniority delta" column matters. A senior leaving for a more senior role at a competitor is a different story than a senior leaving for "Building" with no listed company.
7. Score an inbound deck against the partnership's thesis
Inbound triage. You get the deck, you paste the founder's LinkedIn URL and the one-paragraph summary, the prompt returns a 0-10 across four dimensions and a yes/no recommendation. Calibrated against your fund's own past investments, it gets surprisingly good after the first 30 runs.
This isn't a replacement for the partner read. It's a triage layer so the partner read happens on the right 20%. Most analysts admit they spend two-thirds of their inbound time on decks that were obviously not going anywhere; this prompt is for that two-thirds.
8. Find the warmest path into a target founder
The "I want a meeting" prompt. Cross-references the target's network against your LP and portfolio founder list, returns the strongest mutual ties, and drafts a one-sentence intro request you could send to the top match. No fluff is the right instruction here; the intro request is the deliverable, not a paragraph about why this founder matters.
If the model returns more than one sentence, edit the prompt to be sharper. LLMs are prone to padding intro requests with throat-clearing ("I hope this finds you well, I wanted to reach out about..."), and a VC sending throat-clearing intro requests reads as a junior associate.
9. Daily founder-departure radar across a peer set
The early-morning canary. Runs across 50 series A and B companies in your patch, lists every C-suite or VP departure from the last 24 hours, and flags whether their LinkedIn About now reads like stealth mode.
A daily cadence is the whole game here. Founders coming out of a high- profile role get a short widnow of attention from every fund in their patch. The fund that gets the email at 09:01 the day they leave gets the meeting. The fund that gets it on Friday afternoon doesn't.
10. Surface hiring spikes that signal a fundraise
A 50% increase in open roles weighted toward sales and senior engineering is a strong fundraise signal. Top performers using signal-triggered outreach hit 15-25% reply rates vs 1-2% for cold list, and the same signal works for VCs trying to be in the room before the round closes.
The prompt restricts to functions that telegraph commercial pressure (sales, GTM, senior eng), not just total headcount growth. Engineering hires alone could mean R&D pause; sales-and-eng together is almost always a Series A or B in the next 90 days.
11. Quarterly portfolio activity digest
The partner-meeting prep prompt. Goes through every portco, returns headcount change, new senior hires, new open roles, and any C-suite departure. Formats as a table you paste into the deck.
The "companies to flag" footer is the value-add. Most quarterly portfolio reports say what happened; this one says what to talk about. A portco where the CRO seat has been open for 60 days is the conversation, not the headcount delta.
12. Cofounder-fit shortlist for a portfolio search
Many funds run "platform" or "talent" functions that source senior hires for portfolio companies. This prompt builds a longlist on demand: 30 candidates who match a stage-and-function profile in a target geography, ranked by overlap with the existing founders' background.
Ranked-by-overlap is the trick. A senior IC who worked at the same previous company as the founder, or shares an institutional alma mater, converts to a meeting at roughly twice the rate of a cold longlist match. Optimise for the warm-ish lead, not the perfectly-credentialed stranger.
Frequently asked questions
Does LinkFetch work with my existing dealflow CRM?
LinkFetch is an API and an MCP server. It enriches the data, the CRM holds the records. Most users push LinkFetch output into Affinity, 4Degrees, Folk, or whatever they already run. The prompts above don't replace your CRM; they replace the manual research that fills your CRM.
How does this compare to Crunchbase, Pitchbook, or Specter?
Crunchbase and Pitchbook are firmographic. They tell you what funding a company has raised, when, and from whom. LinkFetch tells you what's happening to the people inside the company right now, who got hired, who left, who changed their About to say "stealth". It's a different layer. Most funds we talk to use both: Crunchbase for the round-stamp, LinkFetch for the leading indicators.
Is it legal for a VC fund to use this kind of LinkedIn data?
Yes, with the same caveats that apply to any vendor in this space. LinkFetch is built on a passive-observer Chrome extension and runs under the user's own LinkedIn session, which is materially different from the fake-account-and-proxies model that got Proxycurl shut down. For the longer compliance answer, see our guide on GDPR and LinkedIn data APIs.
Can I run all 12 prompts on the trial?
Most of them, yes. The 500-credit free trial covers single runs of prompts 1, 4, 6, 7, 8, 11 and a partial run of the others. The cadenced prompts (daily, weekly) are designed to live on a Pro plan; you'd burn the trial in a week running prompt 9 alone.
Do I need to write code to use any of this?
No. Every prompt above is plain English to Claude. The MCP server handles the API calls. If you can paste a prompt into Claude Desktop, you can run the workflow. The optional code path is for funds that want to integrate the same tools into a Slack bot or a CRM webhook, which is documented in the API reference but not required for any of the prompts here.
Last updated: 2026-05-11 by the LinkFetch team.