Set up Claude Code for GTM with one context file

To make Claude Code useful for go-to-market work, give it a CLAUDE.md file that describes your motion, your ICP, your tools and your rules. Claude Code reads that file at the start of every session, so you stop re-explaining your business and start giving it real work.

Most people open Claude Code, see a blinking cursor and have no idea what to type. The problem is rarely the model. It is that the model knows nothing about your company. Without context it behaves like a chatbot. With a well-structured CLAUDE.md it behaves like a GTM operator who has read your playbook.

How CLAUDE.md works

CLAUDE.md is a project memory file. Put it in the root of the folder where you run Claude Code and it is loaded automatically when a session starts. It is plain markdown, so anyone on the team can read and edit it.

Two properties matter for GTM teams. First, it can import other files with an @path/to/file.md line, which lets you keep the root file short and move detail into modules. Second, it lives in the repo, so changes are versioned and shared. When someone updates the ICP, everyone's next session picks it up.

Pick a shape for your motion

A CLAUDE.md for an outbound team looks different from one for a product-led team. Start from the motion you actually run, then add modules. The core question for each is what Claude needs to know to do the daily work well.

If you run more than one motion, keep a minimal root file and let each motion be a module you import.

Build it step by step

1. Create the folder structure. Keep the root clean and give each kind of context its own home. A layout that works for most teams:

   gtm/
     CLAUDE.md
     context/
       icp.md
       signals.md
       voice.md
       exclusions.md
       stack.md
     campaigns/
       log.md
     .claude/
       skills/

2. Write the root CLAUDE.md. Keep it to what Claude needs on every task: who you are, what you sell, the rules that never change and where the rest lives. An example skeleton:

   # Acme GTM workspace
   
   ## Company
   We sell [product] to [buyer role] at [company type].
   Main motion: outbound, supported by founder content.
   
   ## Always
   - Never contact anyone on context/exclusions.md
   - Write in the voice described in context/voice.md
   - Log every campaign change in campaigns/log.md
   - Ask before sending, uploading or deleting anything
   
   ## Context
   @context/icp.md
   @context/signals.md
   @context/voice.md
   @context/exclusions.md
   @context/stack.md

3. Fill in the modules. Each module answers one question. The ICP file says who is a fit and who is not. Signals says which events make an account worth contacting now. Voice holds your tone rules plus two or three real examples. Exclusions lists customers, open deals and competitors. Stack names each tool and what it is used for. If you have not written your ICP down yet, start with the ICP before anything else.
4. Add your first skills. Skills are reusable procedures Claude can pick up when a task matches. Each lives in .claude/skills/<name>/SKILL.md with a name and description at the top. Start with the two tasks you repeat most, for example account research and reply triage.

   ---
   name: account-research
   description: Research a target account and write a one-page brief. Use when asked to research, prep for or qualify a company.
   ---
   1. Read context/icp.md and context/signals.md.
   2. Check the company's site, recent news and open roles.
   3. Score fit against the ICP criteria. Quote the evidence.
   4. Output: summary, fit score, top signal, suggested angle.

5. Connect your tools. Claude Code reaches external systems through MCP servers, added with claude mcp add. Connect read access first (CRM, enrichment) and hold write access until the setup has earned trust. Our MCP servers guide covers the options.
6. Test with a real task. Ask for something you did last week, such as a brief on an account you already know. Compare the output with what you would have written. Where it is wrong, the fix usually belongs in a module, not in the prompt.
7. Keep a campaign log. Ask Claude to append what changed and why whenever a campaign, list or message is edited. Over time this log becomes the memory that stops the team repeating failed experiments.

Modules worth adding later

Once the basics work, most teams add a few more modules as the need shows up:

• Copy frameworks for the email and post structures you trust
• Data hygiene rules: required fields, formatting, what counts as a duplicate
• Experiments listing what is being tested, the hypothesis and the result
• Paid channels with audience definitions, headline rules and approval steps

The same structure scales into a shared company brain. We explain that bigger picture in what a GTM brain is.

Want this built for your team? Book a GTM Engine Review.

Frequently asked

Do I need to be technical to use CLAUDE.md?

No. It is a markdown file you write in plain language. The technical part is installing Claude Code and connecting tools, which is a one-time job.

How long should CLAUDE.md be?

Short enough to read in two minutes. Put detail in imported modules and procedures in skills, and keep the root file for rules that apply to every task.

Can a whole team share one setup?

Yes. Keep it in a git repo so everyone pulls the same context. Personal preferences can go in a local file that is not committed.