Claude Code for GTM. The definitive guide.

The shift

Why Claude Code is taking over GTM workflows

In early 2026, ZoomInfo CEO Henry Schuck built a working enrichment application inside Claude Code in two hours. Not a prototype. A functional app that queried data providers, enriched contacts, and output results. The CEO of the largest B2B data company in the world demonstrated that you no longer need his platform's UI to do what his platform does.

Seth Lee summarized it on LinkedIn: "Claude has been the meta for GTM workflows in 2026." Not "might be" or "could be." Has been.

The shift is not about developers preferring terminals over dashboards. It is about what happens when AI agents become the primary operators of GTM tools.

The old workflow: RevOps person opens Clay. Uploads a CSV. Configures enrichment columns. Waits for results. Downloads CSV. Uploads to HubSpot. Configures a Lemlist sequence. Manually reviews and launches.

The new workflow: "Hey Claude, take this list of 500 leads, enrich them through Apollo and LeadMagic, validate the emails, deduplicate against our HubSpot, and push the qualified ones to our Lemlist outbound sequence."

Same outcome. One is a 45-minute browser workflow. The other is a prompt.

The reason this works in 2026 and did not work in 2024 is tool access. Claude Code can call CLI tools, run scripts, query APIs, read and write files. With Deepline installed as a skill, Claude Code has native access to 30+ GTM data providers. It does not need to "figure out" how to call Apollo's API. Deepline abstracts that away.

What you can build

Five workflows that replace entire tool stacks

1. Waterfall enrichment pipelines

The most common use case. Take a list of contacts, cascade through multiple providers, get the best available data for each person.

# Tell Claude Code:
"Enrich leads.csv with waterfall email enrichment.
Try Apollo first, then LeadMagic, then Prospeo.
Validate all found emails with ZeroBounce.
Output to enriched-leads.csv."

Claude Code runs deepline enrich with the right flags. Handles provider failover automatically. Validates results. Outputs a clean CSV with match rates and cost breakdown.

What this replaces: Clay waterfall columns ($149-349/mo), manual API scripting (hours of dev time), or an SDR manually looking up each contact (hours of human time).

2. ICP scoring from closed-won data

Export your closed-won deals from HubSpot. Ask Claude Code to analyze the patterns: company size, industry, title seniority, tech stack, funding stage. Then score your pipeline against those patterns.

# Tell Claude Code:
"Pull our closed-won deals from HubSpot (last 12 months).
Analyze common patterns: company size, industry, title, tech stack.
Score our current pipeline contacts against those patterns.
Flag the top 50 highest-fit contacts for immediate outreach."

Claude Code queries HubSpot, enriches missing data with Deepline, runs the analysis, and outputs a ranked list. No ML model required. Pattern matching on real data is effective for teams with 100+ closed deals.

3. CRM hygiene automation

Every CRM accumulates garbage data. Duplicate contacts, stale emails, outdated titles, companies that were acquired or went bankrupt.

# Tell Claude Code:
"Audit our HubSpot contacts.
Find duplicates (match on email, then fuzzy match on name + company).
Flag contacts with emails older than 6 months for re-verification.
Re-enrich the flagged contacts through Deepline.
Update HubSpot with fresh data."

This is the kind of project that RevOps teams put on the roadmap and never get to. With Claude Code, it is an afternoon.

4. Signal-based outbound triggers

Monitor for buying signals (job changes, funding rounds, technology adoption, hiring patterns) and automatically trigger outbound when signals fire.

# Tell Claude Code:
"Check for job changes in our target account list (last 7 days).
For any new VP+ hire at a target company, enrich their contact info.
Draft a personalized first-touch email referencing their new role.
Add them to our 'new-exec' Lemlist sequence."

The signal detection, enrichment, personalization, and sequencing all happen in one Claude Code session. No Zapier chain, no manual monitoring.

5. Territory routing with account hierarchy

The hardest GTM data problem: routing leads to the right rep when account hierarchies are complex. Subsidiary of a parent company that is already owned by another rep. Regional offices that map to different territories.

# Tell Claude Code:
"This lead works at Stripe. Check if Stripe or any Stripe subsidiary
is already in our HubSpot as an existing account. If yes, route to
the owning rep. If no, check our territory rules and assign based
on company HQ location and deal size. Update HubSpot ownership."

Claude Code queries HubSpot, resolves the account hierarchy, applies routing logic, and makes the assignment. No custom Salesforce flow required.

Deepline + Claude Code

How the integration works

Deepline ships as a Claude Code skill. When installed, Claude Code can invoke any Deepline command through natural language. No manual API calls or script writing needed.

Installation:

# Install Deepline CLI (skill auto-registers with Claude Code)
bash <(curl -sS https://code.deepline.com/api/v2/cli/install)

# Verify Claude Code can see Deepline
deepline auth status

After installation, Claude Code has access to:

• deepline enrich: waterfall enrichment across 30+ providers
• deepline search: find contacts and companies by criteria
• deepline validate: email verification and phone validation
• deepline dedup: identity resolution and deduplication
• deepline sync: push/pull data from CRMs

What Claude Code sees: Deepline's skill documentation tells Claude Code what commands are available, what parameters they accept, and how to chain them together. Claude Code does not need to read API docs or figure out authentication. The skill handles all of that.

BYOK model: Claude Code uses your API keys (stored locally in your Deepline config). Provider calls go directly from your machine to the provider's API.

Setting up CLAUDE.md

Project instructions for GTM workflows

The CLAUDE.md file in your project root tells Claude Code how to handle your specific GTM context. A well-configured CLAUDE.md turns generic Claude Code into a GTM-aware agent that knows your ICP, your tools, and your rules.

Here is a template for GTM teams:

# GTM Context

## ICP
- Target titles: VP of Sales, Head of Revenue, CRO, VP of Growth
- Company size: 50-500 employees
- Industries: SaaS, FinTech, MarTech
- Funding: Series A through Series C
- Geography: US and Western Europe

## Providers (in priority order for waterfall)
1. Apollo (broadest coverage, free tier)
2. LeadMagic (email verification, pay-per-result)
3. Prospeo (LinkedIn-based email finding)
4. Cognism (European mobile numbers)

## CRM
- HubSpot (main CRM)
- Custom properties: dl_enriched_date, dl_icp_score, dl_email_verified
- Owner mapping: US East = Sarah, US West = Mike, EMEA = Jonas

## Outbound
- Lemlist for email sequences
- Campaign IDs: new-exec-q1 (new hire outreach), re-engage-q1 (dormant accounts)
- Always validate emails before adding to sequences
- Never sequence contacts at existing customer accounts

## Rules
- Always use --dry-run first for batches over 100 contacts
- Log all enrichment costs
- Do not enrich personal email domains (gmail, yahoo, hotmail)
- Re-verify any email older than 90 days before sequencing

With this in CLAUDE.md, you can give Claude Code high-level instructions and it will make the right decisions about providers, routing, and quality checks.

Example end-to-end workflow

From CSV to sequenced outbound in one session

Here is a real workflow that replaces 3-4 tools and 2 hours of manual work.

The prompt:

"I have a CSV at ~/leads/conference-leads.csv with 500 contacts from the SaaStr conference. Enrich them all. Waterfall through Apollo then LeadMagic for email, Cognism for phone. Validate every email. Score them against our ICP (CLAUDE.md has the criteria). Deduplicate against our HubSpot. Push the top 100 highest-scoring leads to our 'saastr-followup' Lemlist sequence."

What Claude Code does:

1. Reads the CSV, identifies columns (name, company, title, LinkedIn URL)
2. Runs deepline enrich --input conference-leads.csv --with '{"alias":"email","tool":"name_and_domain_to_email_waterfall","payload":{"first_name":"{{First Name}}","last_name":"{{Last Name}}","company_name":"{{Company}}","domain":"{{Domain}}"}}'
3. Runs phone enrichment: deepline enrich --input conference-leads.csv --with '{"alias":"phone","tool":"contact_to_phone_waterfall","payload":{"first_name":"{{First Name}}","last_name":"{{Last Name}}","domain":"{{Domain}}","email":"{{Email}}","linkedin_url":"{{LinkedIn URL}}"}}'
4. Validates emails: deepline validate --emails enriched-output.csv
5. Scores contacts against ICP criteria from CLAUDE.md
6. Checks HubSpot for existing contacts: deepline dedup --source hubspot
7. Ranks by ICP score, takes top 100
8. Pushes to Lemlist: deepline sync --target lemlist --campaign saastr-followup

Total time: 10-15 minutes of compute. Zero manual steps after the initial prompt.

Cost transparency: Deepline reports the exact cost at each step. "Apollo: 500 queries, 340 hits, $17.00. LeadMagic: 160 queries (waterfall fallback), 128 hits, $6.40. Cognism: 500 queries, 215 hits, $10.75. Total: $34.15."

No credit system. No surprise bills. No "your plan only includes 3,000 credits and you used 4,200."

When Claude Code is not the right choice

Honest tradeoffs

Claude Code + Deepline is not the right fit for every team. Other approaches win in certain scenarios.

When a visual builder is better: If your RevOps team thinks in spreadsheets and does not use a terminal, forcing them into Claude Code creates friction. Clay or Cargo might be the better fit. The visual feedback loop matters for people who need to see data transforming in real time.

When you need a managed platform: If nobody on your team can debug a CLI error or read a log file, a managed platform with customer support is worth the premium. Deepline assumes a certain technical baseline.

When volume is very low: If you enrich 200 contacts per month, the overhead of setting up Deepline, configuring providers, and writing CLAUDE.md prompts is not justified. Use Apollo's free tier directly.

When compliance is critical: Large enterprises with strict data handling requirements (SOC 2, HIPAA adjacent) might need a vendor with formal compliance certifications. Deepline's BYOK model means data stays on your machine, which is often better for compliance. But some procurement teams want a signed BAA, and that requires a vendor relationship.

The sweet spot for Claude Code + Deepline: teams of 3-50 people doing 1,000-100,000 enrichments per month, with at least one person comfortable in a terminal, who want repeatable workflows and cost transparency.

FAQ

Common questions

$ curl -s "https://code.deepline.com/api/v2/cli/install" | bash