Claude Code for Lead Generation: Find and Qualify Prospects in 2026

Lead generation has always been the bottleneck. Not because finding prospects is impossible — it is because doing it well at scale requires research, enrichment, qualification, and routing. Doing all four manually for 500 companies takes days.

Claude Code changes the arithmetic. As an agentic AI system, it can scrape sources, call enrichment APIs, apply qualification criteria, and output a scored pipeline — without a human in the loop between steps. Teams report compressing 3+ hours of prospect research into under 15 minutes per batch.

This guide covers the full Claude Code lead generation workflow: ICP building, prospect research, enrichment via SyncGTM, lead scoring, and pipeline automation. Every section includes real prompts you can copy and run today.

Overview

This guide is for B2B sales teams, SDRs, RevOps leads, and GTM engineers who want to build a lead generation pipeline with Claude Code — one that runs faster, costs less, and produces better-qualified leads than manual research.

You will learn: how to define an ICP Claude Code can apply automatically, how to run prospect research at scale, how to wire up SyncGTM's enrichment waterfall via MCP, how to score and qualify leads with a repeatable framework, and how to push the results into your CRM or outreach tool automatically.

No code experience required. Every workflow here runs from Claude Code's terminal via natural-language prompts.

Why Claude Code for Lead Generation?

Most teams approach lead generation with a patchwork of tools: one for list building, one for enrichment, one for scoring, one for CRM sync. Each handoff introduces friction, data loss, and human error. Claude Code unifies the workflow into a single agentic session.

Three capabilities make Claude Code particularly well-suited for lead generation:

According to MarketBetter's 2026 analysis, the average task completion time for prospect research drops from 3.1 hours without AI to approximately 15 minutes with Claude Code — a 92% reduction. For teams running weekly prospecting batches, that is 10–12 hours reclaimed per week per rep.

Step 1: Build Your ICP With Claude Code

Ideal Customer Profile (ICP) building is where most lead generation fails. Teams define their ICP once in a slide deck, then ignore it when building lists. Claude Code enforces it automatically — every lead gets scored against the same criteria, every time.

Define your ICP in CLAUDE.md

Store your ICP criteria in your project's CLAUDE.md file. Claude Code reads this at the start of every session — it becomes the permanent context for every lead generation task you run.

## ICP Definition

Target company criteria (score each lead 0–100):
- Industry: B2B SaaS, FinTech, or DevTools (+25 pts)
- Headcount: 50–500 employees (+20 pts)
- Revenue: $5M–$100M ARR (+20 pts)
- Tech stack: HubSpot or Salesforce CRM (+15 pts)
- Funding: Series A or Series B (+10 pts)
- Recent signal: Hired VP Sales/Marketing in last 90 days (+10 pts)

Target persona: VP of Sales, Head of RevOps, or GTM Lead
Exclude: agencies, consulting firms, companies < $2M revenue

Generate your ICP from existing customers

If you have a customer list, Claude Code can analyze it to surface your actual ICP — not the one you assumed you were targeting.

Read customers.csv. It contains our 80 paying customers
with columns: company, industry, headcount, revenue_range,
tech_stack, contract_value, churn_status.

Analyze the data and surface:
1. Top 3 industries by average contract value (not customer count)
2. Headcount range with highest retention (churn_status = "active")
3. Tech stack patterns — which tools appear in 60%+ of accounts
4. Revenue range most correlated with expansion revenue

Output an ICP definition I can paste into CLAUDE.md.

Running this once gives you a data-driven ICP instead of an opinion-driven one. It typically takes 30–60 seconds and surfaces patterns that manual analysis misses.

Step 2: Prospect Research at Scale

Claude Code's agent architecture lets it pull prospect data from multiple structured sources in a single session. Unlike a browser-based AI that forgets context between sessions, Claude Code chains these tasks together automatically.

Sources Claude Code can pull from

Research prompt: company brief in 60 seconds

Research Acme Corp (acmecorp.com). Build a 1-page prospect brief covering:

1. What they do and their primary revenue model
2. Headcount and recent headcount growth (last 12 months)
3. Tech stack — CRM, marketing automation, data tools
4. Recent news: funding, product launches, leadership changes
5. Open roles that signal buying intent for our product
6. Personalization hooks for the VP of Sales persona

Reference our ICP in CLAUDE.md to flag which criteria they hit.
Output as structured markdown.

Bulk research: 50 companies at once

For list-based prospecting, pass a CSV and let Claude Code work through the full list:

Read target_companies.csv. It has 50 company domains.

For each company:
1. Look up headcount, industry, and tech stack
2. Check for recent funding rounds (last 6 months)
3. Check for open VP Sales or RevOps roles
4. Score against our ICP criteria in CLAUDE.md (0–100)

Write results to enriched_companies.csv with columns:
company, domain, headcount, industry, tech_stack,
funding_stage, icp_score, key_signal, notes

Step 3: Enrichment Workflows via SyncGTM

Prospect research tells you which companies to target. Enrichment tells you who to contact and how to reach them. SyncGTM's MCP gives Claude Code direct access to a waterfall of 50+ data providers — so contact enrichment happens inside the same session as research.

What the SyncGTM waterfall covers

Enrichment prompt: contact waterfall

Read prospects.csv. It has columns: first_name, last_name,
company_domain, job_title.

For each row, use SyncGTM MCP to:
1. Run waterfall email enrichment — return verified work email
2. Look up direct-dial phone number
3. Pull LinkedIn profile URL
4. Fetch firmographic data: headcount, revenue_range, industry
5. Check tech stack for HubSpot or Salesforce

Flag any record where email confidence < 80% as "needs review".
Write all results to enriched_prospects.csv.

For deeper context on the enrichment waterfall and how coverage compares across providers, see the guide to waterfall enrichment and how it works. According to G2's lead intelligence category, data coverage and accuracy are the top two purchase criteria for B2B contact enrichment tools — exactly what a waterfall approach addresses.

Step 4: Lead Qualification and Scoring

Qualification is where most manual lead generation breaks down. Reps eyeball a list and guess which leads are hot. Claude Code applies a consistent, documented rubric to every lead — so qualification is auditable and improvable over time.

The points-based ICP scoring model

Define five to eight criteria with explicit point values. Claude Code applies them numerically — no interpretation, no inconsistency.

Routing by score

Scoring prompt

Read enriched_prospects.csv and the ICP criteria in CLAUDE.md.

For each row:
1. Apply the points-based ICP scoring rubric
2. Add a column "icp_score" (0–100)
3. Add a column "routing" (hot / warm / cold)
4. Add a column "top_signal" — the single strongest reason
   this lead scores high (or low)

Sort by icp_score descending.
Write hot leads (score >= 70) to hot_leads.csv.
Write warm leads (score 40–69) to warm_leads.csv.
Write summary stats to console: total, hot count, warm count, cold count,
average icp_score, most common top_signal.

This is the same approach used in the Claude Code sales prospecting workflow — applied here to qualification rather than initial list building.

Step 5: Push Qualified Leads Into the Pipeline

A scored CSV is useful. Qualified leads automatically in your CRM and outreach tool is where the time savings multiply. Claude Code can push leads to most major platforms via their APIs or via MCP connections.

HubSpot: create contacts from hot_leads.csv

Read hot_leads.csv. For each row, create a HubSpot contact
using the HubSpot API (API key in env HUBSPOT_API_KEY):

- Map: first_name, last_name, email, phone, company, job_title
- Set lifecycle stage to "lead"
- Set lead_source to "claude-code-lead-gen"
- Add a note with the top_signal and icp_score
- Skip any row where email is flagged "needs review"

Log success/failure per contact. Output summary at end.

Trigger-based lead generation

The highest-ROI Claude Code lead generation setup monitors for trigger events and runs the full pipeline automatically when they fire. Common triggers:

• Company raises a Series A or B round
• Company posts a VP Sales or Head of RevOps job listing
• A prospect visits your pricing page (via visitor identification)
• A competitor announces a price increase or shutdown
• Contact changes jobs to a target company

For the full trigger-based outreach setup, the sign-up enrichment workflow guide covers how to wire SyncGTM signals into Claude Code sessions.

Claude Code Prompt Library for Lead Generation

Copy-paste these prompts. Each runs as a standalone Claude Code session — no prior setup required beyond having SyncGTM MCP connected and your ICP defined in CLAUDE.md.

Search LinkedIn and Greenhouse for companies currently hiring
a VP of Sales or Head of Revenue Operations.
Target: B2B SaaS companies, 50–500 employees.

For each company found:
- Pull company domain and HQ city
- Check if they appear in our existing CRM export (crm_companies.csv)
- Score against ICP criteria in CLAUDE.md

Output net-new prospects only (not in CRM) to new_prospects.csv.

Read cold_leads.csv — prospects we contacted 6+ months ago
with no response.

For each lead, use SyncGTM MCP to check:
- Has their job title changed?
- Has their company raised funding?
- Are they now hiring for roles in our ICP?

Flag any lead where something has changed as "re-engage now".
Write updated records to reengagement_list.csv with a
"reason_to_reach_out" column summarizing the trigger.

Search G2 and Capterra reviews for companies that have left
reviews for [Competitor Name] mentioning data accuracy,
pricing complaints, or poor support in the last 12 months.

For each reviewer company:
- Look up their domain
- Enrich with SyncGTM MCP (email for decision-maker, tech stack)
- Score against ICP in CLAUDE.md

Output to displacement_prospects.csv.
Include a "pain_point" column with the complaint they mentioned.

For 50 more ready-to-run prompts across research, outreach, and reporting, see the full Claude Code prompt library for sales teams.

Performance Benchmarks: Manual vs Claude Code

Based on data from teams running Claude Code lead generation workflows across B2B SaaS, agency, and RevOps use cases in 2026. For context on how funding signals affect outreach timing, see Crunchbase's research on B2B buying signals— companies that recently raised are 3x more likely to be in active buying mode.

Time savings compound across the pipeline. A team that previously needed 20+ hours/week on list building runs the same volume in under 2 hours — and produces more consistent, better-qualified leads.

For teams wanting to see Claude Code in action across the full sales workflow — not just lead generation — the Claude Code for B2B sales guide covers everything from prospecting through deal close.

Conclusion

Claude Code lead generation is not a toy workflow. Teams running it at scale report 90%+ time reductions across research, enrichment, and qualification — and they get more consistent, auditable results than manual processes produce.

The stack is straightforward. Claude Code handles the agent logic. SyncGTM provides the enrichment data via 50+ waterfall providers. Your CLAUDE.md file holds the ICP criteria that stay consistent across every session.

Start with one workflow — the bulk company research prompt or the contact enrichment prompt. Run it on a list you already have. Measure the time delta. Then wire in qualification scoring and CRM sync.

SyncGTM's free tier covers 100 enrichments per month — enough to validate the workflow before committing to a paid plan. Claude Code Pro is $20/month. The ROI on the first batch of 50 enriched, scored leads typically covers both.