CLAUDE.md, Skills, MCP. GTM setup in 10 minutes.

Who this is for

You do not need to be an engineer

This guide is for RevOps, sales ops, and GTM people who want to use Claude Code for data work. You are not building software. You are enriching contacts, cleaning CRM data, building prospect lists, and automating repetitive tasks.

You do not need to know Python, JavaScript, or any programming language. You need to know what data you want and be specific about describing it in plain English.

Claude Code runs in a terminal. That looks intimidating if you have never used one. But the actual interaction is conversational - you type what you want in natural language, and Claude Code executes it. The terminal is just the interface. The real work is knowing what to ask for.

Three things make Claude Code significantly better for GTM work:

1. CLAUDE.md - a file that tells Claude Code about your project, your ICP, your tools, your rules
2. Skills - pre-built instruction sets that teach Claude Code specific capabilities
3. MCP - connections to external tools and APIs

Set up all three and Claude Code becomes a GTM co-pilot that knows your business.

CLAUDE.md

What CLAUDE.md is and why it matters

CLAUDE.md is a markdown file you put in your project folder. Claude Code reads it automatically at the start of every session. Think of it as onboarding documentation for an AI teammate.

Without CLAUDE.md, every Claude Code session starts from zero. You have to explain your ICP, your preferred tools, your workflow rules, every single time. With CLAUDE.md, Claude Code already knows all of this before you type your first prompt.

Your first CLAUDE.md for GTM

Create a file called CLAUDE.md in your project folder. Here is a practical starting template:

# GTM Project

## Our ICP
- Series B-D SaaS companies
- 50-500 employees
- US, UK, or EU based
- Must have a sales team of 5+ people

## Tools
- Deepline CLI for enrichment (waterfall order: apollo -> leadmagic -> prospeo)
- Email validation: zerobounce (primary), neverbounce (fallback)
- Company data: crustdata (primary), pdl (fallback)
- CRM: HubSpot
- Outbound: Lemlist for sequences

## Rules
- Always validate emails before adding to any sequence
- Deduplicate against HubSpot export before enrichment to avoid paying for data we already have
- Use BYOK mode (our API keys, not Deepline hosted keys)
- When enriching, waterfall cheapest provider first
- Flag any contact without a valid email - do not add to sequences

## Output format
- All outputs as CSV
- Include provider_used column so we can track which providers are hitting
- Sort enriched lists by lead_score descending

That is it. Save this file. Next time you open Claude Code in this folder, it already knows:

• What kind of companies you target
• Which enrichment providers to use and in what order
• Your validation rules
• Your CRM and sequencing tools
• How you want output formatted

What to put in CLAUDE.md vs what to say in prompts

Put in CLAUDE.md: Things that do not change between sessions. ICP definition, provider preferences, validation rules, output format standards, tool configurations.

Say in prompts: Things specific to this task. "Enrich this list." "Find contacts at these 20 companies." "Build me a prospect list for the healthcare vertical."

The split is: CLAUDE.md handles the "how" (rules, tools, preferences). Your prompts handle the "what" (this specific task, this specific data).

Iterating on your CLAUDE.md

Your CLAUDE.md will evolve. After a few sessions, you will notice patterns:

• "I keep telling Claude to sort by score" - add it to CLAUDE.md
• "I always want the LinkedIn URL column" - add it to output format rules
• "This provider is not working well for European contacts" - update the waterfall order

Treat CLAUDE.md like a living document. Update it after every session where you had to repeat yourself.

Skills

Skills: pre-built knowledge for Claude Code

A Skill is a set of instructions that teaches Claude Code how to do something specific. Instead of explaining enrichment workflows every time, you install the Deepline skill and Claude Code already knows how to run waterfall enrichment, validate emails, build prospect lists, and more.

Installing Deepline as a Claude Code skill

The Deepline CLI ships with a Claude Code skill built in. Install it:

bash <(curl -sS https://code.deepline.com/api/v2/cli/install)

Then verify the installation:

deepline auth status

Once installed, Claude Code can:

• Run waterfall enrichment across 30+ providers
• Validate emails before outbound
• Enrich companies with firmographic data
• Build prospect lists from ICP criteria
• Deduplicate and clean contact data
• Track costs and provider performance

You do not need to explain any of this in your prompts. The skill teaches Claude Code the commands, the flags, the output formats, and the error handling.

What happens when you combine CLAUDE.md + Skills

CLAUDE.md says: "Our ICP is Series B SaaS, 50-500 employees, US-based. Waterfall through apollo then leadmagic then prospeo."

The Deepline skill says: "Here is how to run waterfall enrichment, here are the available providers, here is how to handle failures."

Your prompt says: "Build me a prospect list of 30 companies. Find VP of Sales at each."

Claude Code combines all three and executes the complete workflow: finding companies, enriching contacts, waterfalling for emails, validating, scoring, and outputting a sorted CSV. You asked one question. The layers of context made the answer complete.

Other useful skills for GTM

Deepline is the core data skill. But Claude Code can also work with:

• HubSpot MCP: read and write CRM data directly
• Spreadsheet tools: manipulate CSVs and Excel files
• Web browsing: research companies, check job postings, verify information

Each skill you add increases what Claude Code can do without detailed instructions from you.

MCP

MCP: connecting Claude Code to external tools

MCP (Model Context Protocol) is a standard for connecting AI tools to external APIs and data sources. In practical terms: it lets Claude Code call APIs without you writing code or even knowing what an API is.

Why MCP matters for GTM

Without MCP, the workflow for getting data from an external tool is:

1. Open the tool (HubSpot, Apollo, etc.)
2. Export data manually
3. Give the CSV to Claude Code
4. Claude Code processes it
5. You manually import the result back

With MCP:

1. Tell Claude Code what you want
2. Claude Code calls the tool directly, processes data, and pushes results back

The CSV export/import cycle disappears. Claude Code talks to your tools directly.

Deepline MCP integration

Deepline's MCP server means Claude Code can call enrichment APIs as part of any workflow. Instead of running a CLI command and waiting for a CSV, Claude Code can enrich data inline while building a prospect list, writing personalized emails, or updating your CRM.

The practical difference is speed and context. Claude Code can enrich a contact, use that enrichment data to write a personalized email, and push both to your CRM - all in one continuous workflow. No intermediate files. No copy-pasting between steps.

Setting up MCP connections

MCP servers are configured in your Claude Code settings. The Deepline CLI installer handles its own MCP configuration. For other tools:

HubSpot MCP: Connects Claude Code to your CRM for reading contacts, companies, deals, and writing updates back.

Browser MCP: Lets Claude Code visit websites to research companies, check job postings, and verify information.

The setup for each is typically a one-time configuration with an API key or OAuth token. Once connected, Claude Code can use these tools in any session automatically.

Practical walkthrough

From zero to running GTM workflow

Here is the step-by-step setup, assuming you are starting fresh.

Step 1: Install Claude Code

Download from claude.ai/code. Requires a Claude Pro or Team subscription.

Step 2: Install Deepline CLI

Open your terminal (or Claude Code) and run:

bash <(curl -sS https://code.deepline.com/api/v2/cli/install)

This installs the Deepline CLI and the Claude Code skill. Verify with:

deepline auth status

Step 3: Create your project folder and CLAUDE.md

Create a folder for your GTM work. Inside it, create a CLAUDE.md file with your ICP, tools, and rules (use the template from the CLAUDE.md section above).

mkdir ~/gtm-workspace
cd ~/gtm-workspace

Create CLAUDE.md with your favorite text editor, or tell Claude Code to create it for you:

Create a CLAUDE.md for our GTM project. We target Series B-D SaaS companies,
50-500 employees, US-based. We use Deepline for enrichment with apollo,
leadmagic, and prospeo as waterfall providers. HubSpot for CRM. Lemlist
for outbound.

Step 4: Test with a simple enrichment

Put a small CSV (5-10 rows) in your project folder. Then open Claude Code and ask:

Enrich leads.csv and find the VP of Sales at each company. Waterfall for emails.
Validate all emails before output.

Watch what happens. Claude Code reads your CLAUDE.md, uses the Deepline skill, runs the enrichment, validates emails, and outputs a clean CSV. All from one prompt.

Step 5: Build from there

Once the basic workflow works, layer on complexity:

• "Every Monday, re-verify emails in my active sequences"
• "When I paste a company URL, find 5 contacts matching our ICP"
• "Score this list against our closed-won data"
• "Build a prospect list of 50 companies in fintech, then find decision makers and write personalized first-touch emails"

Each of these is one prompt. CLAUDE.md + Skills handle the context. You handle the intent.

Common patterns that work

Workflows GTM teams run every week

The Monday pipeline refresh

Add to your CLAUDE.md:

## Weekly Monday Workflow
1. Export active pipeline from HubSpot
2. Re-validate all emails using zerobounce
3. Re-enrich any contacts missing phone numbers
4. Flag contacts whose company changed (job moves, funding rounds)
5. Output refreshed_pipeline.csv and changes_log.csv

Then every Monday, open Claude Code and say: "Run the Monday refresh." That is the entire prompt. CLAUDE.md tells it what to do.

The instant research workflow

When a prospect books a meeting:

Research acme.io. Find their tech stack, recent funding, headcount growth,
and 3 recent news items. Also find 2 other contacts at the company who
might be in the buying group. Output a one-page brief.

Claude Code uses Deepline for enrichment data and web browsing for news. You get a prospect brief in under a minute that would take 20 minutes of manual research.

The competitive displacement workflow

Find 50 companies using Outreach.io using Deepline with builtwith. Then find the
VP of Sales or Head of RevOps at each. Waterfall emails. Score by company size
and validate all emails. Output to outreach_displacement.csv.

Signal-based prospecting with one prompt. The enrichment data is real. The contacts are verified. The list is ready for your sequencing tool.

The CRM hygiene workflow

Read crm_export.csv. Find duplicates by email and domain. Normalize all job titles.
Flag contacts with invalid or catch-all emails. Output clean_crm.csv and a
changes_log.csv showing every modification.

Run this monthly. Clean CRM data means better lead routing, better reporting, and better deliverability.

FAQ

Frequently asked questions

Related

Keep reading

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