Claude Code for Investors: Deal Sourcing, Market Maps

Claude Code for investors means running deal sourcing from one terminal. You write the investment thesis as filters (sector, latest round, headcount band, geography, growth), Claude Code queries a company and people database, and you get back a sourcing list, a market map and a watchlist of portfolio and competitor signals as files you can check. Each row shows the fields that put it there.

Proactive sourcing is already a large share of how venture deals start. In the survey of 885 venture capitalists by Gompers, Gornall, Kaplan and Strebulaev, summarized on the Harvard Law School Forum on Corporate Governance (2019), almost 30% of deals were proactively self-generated and only 10% came inbound from company management.

This guide covers eight steps in plain language, a worked example built on a live search, which steps cost credits, the metrics to track, the honest limits and the common mistakes. It is written for VC and PE investors, platform and scout teams, and corporate development.

Quick answer: To use Claude Code for investors, connect a company-data layer such as the Landbase CLI, write the thesis as filters, build a sourcing list with a free default search, rank it by hiring and headcount signals, group it into a market map with written definitions, monitor portfolio and competitor leadership changes, find founders with LinkedIn URLs, then push the shortlist to your CRM or Affinity.

Before you start: what you need

A sourcing run is only as precise as the thesis you give it. Have these ready before the first prompt.

• Claude Code installed, with one project folder per thesis and a short CLAUDE.md holding the thesis filters, exclusions and review rules. Claude Code loads CLAUDE.md into every session, per the Claude Code memory docs.
• The Landbase CLI installed and signed in (Step 1).
• A written thesis: sector, the latest rounds you invest in, a headcount band, geographies, and what counts as "growing" for you.
• A CSV of current portfolio companies and named competitors, with a company name and website per row, for monitoring and exclusion.
• An export of companies already in your pipeline (Affinity, HubSpot, Salesforce or a spreadsheet), so the list only shows new names.
• A named reviewer, usually an associate or platform lead, and a rule for how many rows get checked by hand before anyone sends an intro request.

How to use Claude Code for investors: step by step

Step 1: Connect Landbase to Claude Code

Install the CLI with the install and connect guide, sign in, and run the built-in health check to confirm setup, sign-in and connectivity. Then start each Claude Code session by loading Landbase into it. The setup is the same one used for Claude Code sales prospecting and Claude Code recruiting. You are done when a test search returns a list. Cost: none.

Step 2: Write the thesis as filters

Turn the thesis memo into five filters and mark each one hard or soft.

• Sector: "cybersecurity companies that sell to enterprises", plus the categories you exclude (services firms, resellers).
• Stage: the latest recorded round, for example Series A or Series B.
• Headcount band: 20-200 employees.
• Geography: headquartered in the United States.
• Growth: at least one open job posting now, and headcount up over a period you name (for example 90 days).

Sector, stage, size and geography go into the search. Growth is a ranking signal you add in Step 4, because headcount trend is not a default search filter. A good first request is to have Claude Code split the thesis into search filters and ranking rules.

Step 3: Build the sourcing list

Ask Claude Code for the list in plain English, for example Series A and B US cybersecurity companies with 20-200 employees. Keep the work in one named session, so later turns refine the same list.

The reply includes the list, a funnel of counts and the filters the agent chose. Read the filters before you trust the count; sector words often map to broad industry codes. To widen from companies you already like, ask for lookalikes of two of your best-performing portfolio companies. For a larger expansion, Claude Code can grow a seed list into similar companies with the same geography and stage rules. Decision rule: under 50 companies, widen the size band or geography; over 3,000, tighten the sector definition. Cost: none for a standard search.

Step 4: Rank by hiring and headcount momentum

Each company row carries total headcount, department headcounts such as engineering and sales, and the number of active job postings. Ask Claude Code to rank the list by open postings and show a free preview before saving it.

Headcount growth needs two points in time. The free way is to save the list today, rerun the same search in 30 or 90 days and have Claude Code compare the two snapshots company by company. The faster way is a research-mode query (for example, companies that grew headcount more than 20% in the last 6 months): the first turn drafts the query and quotes credits for free, the second spends them, so approve the second turn only after review. A sales headcount rising faster than engineering often marks a go-to-market push; the hiring signals guide covers role-level postings. Cost: snapshots none; the second research-mode turn spends credits.

Step 5: Build a market map with stated definitions

A market map is only useful if a partner can see why each company sits in its box. Write the category rules first, then let Claude Code apply them to every row:

1. Define each category in one line, for example "Identity: description or keywords mention identity, access or authentication".
2. Define regions the same way, for example "New England: HQ in MA, CT, RI, NH, VT or ME".
3. Have Claude Code assign one category and one region per row, with an explicit "unclassified" bucket.
4. Preview the count of companies in each category and region before you save the map.

Keep the definitions in the project folder next to the map, with the run date. If more than a third of rows land in "unclassified", the rules are too narrow. For fuzzy categories, AI qualification answers a yes/no question per row from the company description. Cost: sorting and previews none; qualification uses credits per row, so ask for a credit estimate first. The TAM mapping guide covers sizing in more depth.

Step 6: Monitor portfolio and competitors

Give Claude Code the portfolio and competitor CSV and have it prepare the list for monitoring:

1. Upload the CSV as a new list.
2. Clean up the names and websites, then match each row to a company record.
3. Add the fields you want to watch, such as headcount and active job postings. A field the database does not recognize comes back empty without an error, so preview the columns before you rely on them.

For leadership changes, ask for the C-level executives and VPs at the companies on that list. Each person row shows management level and months in the current role, so anyone at C-level with six months or less is a new leader to note. Rerun monthly and compare against last month. Flag any company whose headcount moves more than 10% in a quarter, or that adds a CFO, CRO or first VP of Sales. Cost: none.

Step 7: Find founders and operators

For the shortlist, ask for the founders and CEOs at each company. Each row includes a LinkedIn URL, so the associate can read the profile and check for a warm path before outreach. For operator networks (advisors, board candidates, executives for a portfolio company), search by current role, for example VPs of Sales at Series B cybersecurity companies in New York. If you will email rather than message on LinkedIn, enrich only the final rows with email addresses. Cost: searches none; contact enrichment uses credits per contact, so ask for a credit estimate first.

Step 8: Push to the pipeline and set the cadence

For HubSpot or Salesforce, have Claude Code drop companies already in the CRM first, then create records for the rest under a named list such as "Cyber A-B Nov 2026". The push only adds new records, so existing records stay untouched. For Affinity, export the list as a CSV and import it into a list. Affinity matches organizations on website and requires a name, per its import guide. Tag every row with the thesis and run date. Cadence: signals weekly, snapshots monthly, thesis filters quarterly. Cost: none.

Worked example: Series A and B cybersecurity in the United States

The counts below come from one standard Landbase search run on November 3, 2026, plus free previews of the same list. The last two bullets are an illustrative example.

• Search: the Step 3 query returned 1,852 United States companies with 20-200 employees whose latest recorded round is Series A (1,178) or Series B (674).
• Definition check: 1,653 of those rows sat under the broad "Software Development" industry code. A free filter keeping rows in a security industry code, or whose description or keywords mention security, left 458 companies with a median of 53 employees.
• Momentum: 173 of the 458 had at least one active job posting, and 85 had five or more.
• Market map by region: California 170, New York 75, Texas 29, New England 28, DC metro 28, other states 103, HQ state blank 25.
• Market map by category: simple keyword rules placed 82 in identity, 48 in cloud security, 36 in data security, 13 in email security and 11 in application security; 268 stayed unclassified, a sign the rules need a qualification pass.
• Leadership: of 197 C-level contacts attached to these companies, 8 had been in role six months or less.
• Shortlist (illustrative example): the 85 companies with five or more postings go to qualification, about 60 pass the "sells security software to enterprises" question, and the associate reviews 25 by hand.
• Pipeline (illustrative example): 20 new names are pushed to the CRM with the thesis tag, and founders' LinkedIn URLs go to the partner who owns the category.

The free steps cut 1,852 rows to 85 before any credit was spent. Running filters before paid steps is the main cost control in this workflow.

How to measure whether it worked

• List precision: the share of a 25-row hand-checked sample that truly fits the thesis. Hold outreach until it reaches 80%.
• Coverage of known deals: of the last 20 companies in your thesis that raised or that you met, how many appear in the list. Low coverage means the filters are too narrow.
• Proactive share of pipeline: new pipeline entries tagged as self-sourced, compared against the roughly 30% proactive baseline in the Gompers et al. survey.
• Signal-to-first-touch time: days from a flagged signal (new CRO, posting spike) to a first meeting request.
• Credits per qualified company: qualification and enrichment credits divided by companies that reach first meeting.

Review the numbers monthly at the pipeline meeting, and rerun the full list whenever the thesis changes.

The honest limits of Claude Code for investors

• The funding field is the latest recorded round only. It shows one round, not the cap table or round history, and in the example run the search agent reported that funding stage is populated for under 2% of company records globally. Companies with no recorded round drop out of a stage filter. Confirm the round, amount and date on the company's announcement or filings.
• Headcount from profiles differs from press figures. Employee counts are built from people who list the company, much like LinkedIn's Page count, which it describes as the number of employees linked with the Page. Contractors, offshore teams and people without profiles are missed, so treat counts as a trend, not the company's reported number.
• Verify before outreach. Titles and roles change, and data lags; the job change frequency statistics show how often. Check every founder's profile before an intro request.
• Web and news are still better for fresh announcements. A round announced this week, an acquisition or a founder departure shows up in news before it reaches a database. Pair the list with a news alert or the search tool's web lookup for anything time-sensitive.
• Analysis is not a decision. Gartner predicted in 2021 that by 2025 more than 75% of VC and early-stage executive reviews would be informed by AI and data analytics. Informed is the key word: the list decides who gets a look, and the partner still decides who gets a term sheet.

Common Claude Code deal sourcing mistakes

Trusting the sector label

Searches feel precise when a count comes back fast. In the example, most rows sat under a generic software code, and a market map built on that label would have been mostly noise. Write the sector definition as a rule, test it on a 25-row sample, and keep the rule in the project folder.

Using stage as a hard filter without a fallback

Funding coverage is sparse, so a strict stage filter hides companies that raised quietly or recently. Run a second pass without the stage filter, ranked by headcount and hiring, and review the top rows by hand.

Reading one headcount snapshot as growth

A company with 150 employees is not growing because it is large. Growth needs two dated snapshots or a research-mode query; label any single-point number as size, not momentum.

Pushing unreviewed rows into the pipeline

Bulk pushes fill Affinity or the CRM with names nobody owns, and partners stop trusting the list. Push only the reviewed shortlist, tagged with thesis, run date and owner. For ranked company examples by investor, see the list of fastest-growing companies backed by Founders Fund.

How to do this in Landbase

Landbase covers 40M+ companies and 800M+ contacts from 20+ providers, with 1,500+ fields for enrichment. From Claude Code, the CLI turns a plain-language thesis into a company list, ranks it with free sorting and previews, and charges credits only for qualification, contact enrichment and research-mode queries, each quoted before it runs. New accounts start with 1,000 free credits. The same workflow powers Claude Code marketing workflows and RevOps automation.

Frequently Asked Questions

Can Claude Code do VC deal sourcing on its own?

Not on its own. Claude Code can plan queries, sort data and organize files, but it needs a data source to return real companies. Connected to a company database through a CLI, it can build a sourcing list, rank it and push it to a CRM in one session. Without one, it only knows what is on the web or in its training data.

What are the best VC deal sourcing tools to pair with Claude Code?

Most teams combine three kinds of tools: a company and people database for lists and signals, a relationship CRM such as Affinity, HubSpot or Salesforce for pipeline, and a news or funding feed for fresh announcements. Claude Code sits on top and runs the steps between them. Pick the data layer by coverage of your stage and geography, and test it against deals you already know.

How accurate is headcount data for tracking startup growth?

It is reliable as a trend and less reliable as an exact number. Counts built from professional profiles miss contractors and people without profiles, so they often differ from what a company tells the press. Compare the same source over time, and treat moves above 10% in a quarter as worth a look rather than as proof.

How much does AI deal sourcing cost in credits?

Searching, filtering, ranking, market maps and CRM pushes in this workflow cost no credits. Credits are spent on AI qualification per row, contact enrichment per contact, and the second turn of a research-mode query. Ask for a credit estimate before each paid step, and run the free filters first so paid steps touch fewer rows.