There's a wave of content right now telling GTM teams to replace their entire stack with Claude Code. Build everything from scratch. Skip the SaaS tools. Vibe your way to pipeline.

That's not the right framing.

Claude Code is genuinely powerful for GTM work. But like any tool, it's most useful when you understand exactly what it does well, what it doesn't, and how it fits alongside everything else you're already running.

At Partner UP, we use Claude Code and Skills on client projects. Not to replace Clay or HubSpot, but for specific workflows where a terminal-based agent with full filesystem access and the ability to call APIs in sequence does things those tools can't.

This post covers the practical reality: what Claude Code and Skills are, what they're good for in outbound and RevOps specifically, and where you still need purpose-built tooling.

What Claude Code Actually Is

Claude Code is a terminal-based AI agent. You open a terminal, type claude, and interact with an agent that can read and write files on your machine, execute code, call external APIs, and chain multi-step workflows together, all from plain English instructions.

That last part is what separates it from chat-based AI tools. Chat tools generate text you then copy and paste somewhere. Claude Code executes actions. It can pull a CSV from your project folder, enrich each row with API calls, filter by ICP criteria, and output a formatted prospect list as a single workflow, without you writing a script.

For GTM Engineers and RevOps practitioners, that distinction matters because it means workflows that previously needed a developer — lead sourcing scripts, CRM data cleanup pipelines, signal processing automations — are now afternoon projects.

The honest caveat: Claude Code works locally. It has no persistent scheduling, no native CRM connectors, and no built-in team collaboration. Workflows you build run on your machine when you trigger them. That's a real limitation, and it shapes where in the stack it belongs.

If you want a structured introduction to how Claude Code works in practice, Anthropic's Claude Code in Action course covers the core development workflow from setup through real use cases.

What Skills Are and Why They Matter for GTM

A Skill is a folder in your Claude Code project that contains a SKILL.md file — a set of instructions Claude reads automatically when it detects a relevant task.

Without Skills, every Claude Code session starts from scratch. You re-explain your ICP. You re-describe your data format. You re-specify your outreach style. That overhead compounds across dozens of sessions.

With Skills, that context is permanent. You define it once in a SKILL.md file, and Claude applies it every time automatically. For GTM work specifically, this means:

• Your ICP criteria are embedded in every prospecting workflow without re-stating them

• Your outreach voice and style apply to every copy generation task by default

• Your CRM field structure is understood before you ask for a cleanup script

• Your competitor positioning is baked into every research task

Skills are just files in a project directory. If your project lives in a Git repo, every team member who pulls the repo gets the same Skills automatically. When one person improves a Skill, the whole team benefits on their next session.

This is the part most Claude Code content skips. The tool itself is powerful. The Skills layer is what makes it a system rather than a collection of one-off prompts.

Anthropic's Introduction to agent skills course covers how to build, configure, and share Skills, including how to distribute them across a team through a shared repo.

Claude Code for Outbound and Sales

These are the workflows where Claude Code earns its place in a sales GTM stack.

ICP analysis from CRM data

Most ICP definitions are built on assumptions rather than data. Claude Code changes that. Export your closed-won and closed-lost deals from HubSpot as a CSV, drop it into the project folder, and ask Claude to find which segments actually convert at the highest rate: by industry, company size, tech stack, funding stage, or whatever fields you have.

The output is an ICP built on your own deal history, not a persona template from a blog post. Run it quarterly as new deal data accumulates.

Signal-based lead sourcing

Give Claude Code a target account list and a set of buying signals — recent funding rounds, specific job postings, tech stack indicators, LinkedIn activity — and it builds a sourcing workflow that pulls matching companies from public sources and APIs. The workflow outputs a structured list with the signals that triggered each account.

This is where Claude Code and Clay genuinely complement each other. Clay is better for visual, no-code enrichment pipelines that run continuously and sync to your CRM in real time. Claude Code is better for the initial research and logic design: figuring out which signals actually predict conversion in your segment, then building the filter criteria that Clay runs at scale. See our GTM partners page for the tools we use alongside both.

Personalization at scale

Build a Skill that defines your outreach voice, your ICP, and your value proposition per segment. Then feed it an enriched prospect list. Claude Code generates first lines or full email drafts for each contact, drawing on the specific signals in the data: a recent funding announcement, a job posting that implies a pain point, a LinkedIn post relevant to your offer.

The quality depends entirely on the enrichment quality going in. Weak data produces generic personalization. This is why the enrichment step comes first.

Pre-call research briefs

Before a discovery call, a rep needs context: the company's recent news, the contact's background, the account's current tech stack, any previous interactions in the CRM. Claude Code can pull all of this in under two minutes from a terminal command. Build a Skill for it, and any rep can trigger it with a single instruction before any call.

Claude Code for RevOps

RevOps work involves a lot of tasks that are structured enough to describe clearly but messy enough that rigid automation tools struggle with them. Claude Code handles the messy middle well.

CRM data cleanup

Export your HubSpot contacts or deals as a CSV. Describe the problem in plain English: duplicate records, inconsistent company name formats, missing lifecycle stages, job titles that need standardization. Claude Code reads the file, identifies the issues, writes a cleanup script, and outputs a corrected version for you to review before importing back.

A cleanup script that would take a RevOps analyst several hours to build manually gets generated in minutes. The key is that you review the output before it touches your CRM. Claude Code is good at identifying patterns and writing the logic. A human should verify before production data changes.

Pipeline and forecast analysis

Export your open pipeline from HubSpot. Ask Claude Code to identify deals that haven't moved in 30 days, flag opportunities with missing qualification fields, calculate weighted pipeline by stage, and surface which rep's pipeline has the weakest stage distribution. The output is a structured analysis you can use directly in a pipeline review.

This doesn't replace your HubSpot reporting. It supplements it for ad hoc analysis where building a custom report in HubSpot would take longer than the insight is worth.

Workflow documentation

One of the most underrated RevOps use cases. If you've built HubSpot workflows over time without proper documentation, Claude Code can read your exported workflow logic and generate structured documentation for each one: what triggers it, what it does, what it affects downstream. This matters for onboarding new team members and for auditing automation before it breaks something.

Custom CRM scripts and integrations

When you need something HubSpot's native functionality doesn't cover — a custom data transformation, a one-off sync between two tools, a script that processes a specific data format — Claude Code builds it from a plain English description. You don't need a developer. You need to be able to describe what you want clearly and review the output critically.

The limitation worth repeating: these scripts run locally when you trigger them. For workflows that need to run automatically on a schedule or be accessible to a whole team, you'll move the logic into n8n, Make, or HubSpot workflows once it's validated.

For a closer look at how Partner UP approaches RevOps implementation and GTM strategy, those pages cover the broader system these tools fit into.

Claude Code vs. Clay: How They Actually Fit Together

The question that comes up most often is whether Claude Code replaces Clay. It doesn't. They solve different problems and they work better together than either does alone.

The practical split we use at Partner UP: Claude Code for the design and validation phase, Clay for the execution phase at scale. You figure out which signals predict conversion and what enrichment logic works in Claude Code. Then you build the ongoing workflow in Clay where it can run continuously and sync to HubSpot automatically.

If you're not sure where Claude Code fits in your current stack, the answer usually depends on where your biggest manual bottlenecks are. Partner UP works with GTM and RevOps teams to map the right tool to the right workflow.

Setting It Up: The Practical Minimum

Getting started with Claude Code for GTM doesn't require much. Here's the minimum viable setup.

1. Install Claude Code

Claude Code requires Node.js. Install it from the terminal:

npm install -g

npm install -g

npm install -g

You'll need a paid Anthropic account (Pro at $20/month or Max at $100/month for heavier usage). Log in when prompted after installation.

2. Create a project folder

Every GTM project should have its own folder. This is where your data files, output files, and Skills live. Keep your CRM exports, prospect lists, and reference documents here.

3. Build a CLAUDE.md context file

This is the most important setup step. Create a CLAUDE.md file in your project folder and define:

• Your ICP: company size, industry, geography, tech stack indicators, exclusion criteria

• Your outreach voice: tone, sentence length, what to avoid, example copy

• Your CRM structure: key fields, stage definitions, property naming conventions

• Your tools and integrations: what APIs are available, what data sources to use

Claude reads this file automatically at the start of every session. The upfront investment is about 30 minutes. It saves that time back on every subsequent session.

4. Add Skills for repeatable workflows

For any workflow you'll run more than once, create a Skill. Start with the highest-frequency tasks: pre-call research, personalization, CRM cleanup. Each Skill is a folder with a SKILL.md file that describes what the workflow does, what inputs it expects, and what output it produces.

Once your core Skills are in place, you can trigger complex multi-step workflows with a single plain-English instruction. Claude selects the right Skills, chains them in order, and passes context between steps automatically.

For more complex chaining and delegation, Anthropic's Introduction to subagents course covers how to use subagents to manage context across longer workflows, useful when a GTM workflow spans multiple tools and data sources.

Where the Limits Are

Claude Code is worth using. It's also worth being clear-eyed about what it doesn't do.

No persistent scheduling. Workflows run when you trigger them, not on a cron job. For anything that needs to run daily or hourly without human input, use n8n or HubSpot workflows.

No native CRM sync. Data moves through exports and imports. Real-time bidirectional sync with HubSpot or Salesforce still needs purpose-built connectors.

No waterfall enrichment. Building fallback logic across multiple data providers in Claude Code means managing API keys and writing custom error handling. Clay handles this out of the box.

Local execution. If your team needs to run shared workflows, the scripts need to be moved to a shared environment. Claude Code itself doesn't have team collaboration features built in.

Output quality depends on input quality. Weak data, vague ICP definitions, or poorly written Skills produce inconsistent output. The garbage-in principle applies here more than anywhere.

None of these are reasons not to use it. They're reasons to use it for the right things.

Common Questions

Do I need to know how to code to use Claude Code for GTM?

No. The workflows in this post are driven by plain English instructions. You describe what you want, Claude Code figures out how to do it. Where you do need some technical comfort is in reviewing outputs, especially scripts that will touch your CRM data. You don't need to write the code, but you should be able to read it well enough to catch errors before they run on production data.

Is Claude Code better than n8n or Make for GTM automation?

Different tools, different jobs. n8n and Make are better for trigger-based workflows that run automatically on a schedule and connect SaaS tools. Claude Code is better for research-heavy, judgment-intensive tasks that are hard to encode as rigid IF/THEN logic. Most mature GTM stacks use both: Claude Code for design and one-off analysis, n8n or Make for productionised automation.

How do Skills relate to Claude's memory?

Claude Code has no memory between sessions by default. Skills and context files (CLAUDE.md) are the workaround. They're just files Claude reads at the start of each session, giving it the context it needs without you re-explaining everything. Think of them as your standing operating procedures for the agent.

If you're newer to Claude generally, Claude 101 covers the core features and how to get the most out of Claude before layering in Code and Skills.

Can Claude Code replace a GTM Engineer?

No. It replaces the repetitive, data-heavy parts of the job: manual research, data processing, template management, context assembly. It amplifies the judgment-intensive parts: ICP design, signal selection, messaging strategy, system architecture. The GTM Engineers who learn to use it well get dramatically more done. The work doesn't disappear; it shifts toward higher-value decisions.

The Tools That Matter Are the Ones You Actually Use

Claude Code and Skills are worth learning if you're doing serious GTM or RevOps work. The learning curve is real but short. The productivity gain, once your Skills and context files are in place, is significant.

What it isn't is a replacement for the system design work that comes before it. You still need a clear ICP, a validated motion, and clean data. Claude Code makes a good system faster. It doesn't substitute for one.

Partner UP works with GTM and RevOps teams on exactly this layer: designing the system and building the right tools into it. If you're figuring out where Claude Code fits in your stack, that's a good conversation to have.

Written by Leila Ergul Demir, Founder of Partner UP. Leila helps companies design and implement scalable GTM systems and revenue operations. She specializes in helping founders, revenue leaders, and RevOps teams.