ShipFit
Get Started
Integration

Claude Code Project Setup: Ship the Validated Playbook into Your Terminal

Stage 9 of the ShipFit playbook ships your validated decisions to Claude Code so the CLI agent works from your buyer, V1 scope, pricing, and launch plan from the first prompt.

What Claude Code does

Claude Code is Anthropic’s terminal-native coding agent. It reads your repo, writes code, runs tests, and commits, all from the command line. Like Cursor, it leans on Claude as the underlying model. Unlike Cursor, it lives in the terminal instead of an IDE, which makes it the default choice for backend, CLI, or infra work where the IDE is overhead.

Anthropic’s documentation describes how Claude Code picks up project context on session start; once that context is loaded, anything in it persists across every prompt for the lifetime of the project. That makes Claude Code the cheapest place to keep your validation decisions in the agent’s hands.

What ShipFit ships to Claude Code

When you reach stage 9 (What to Export?) and pick Claude Code, the export takes the validated playbook produced by the previous 8 stages and lands it in your repo as a Claude Code project. The agent then works from your validated decisions, not generic defaults.

What carries over from the playbook into the export:

  1. Buyer persona from stage 2 (Who Pays?), including the verbatim quotes from your interviews.
  2. Above-the-line pains from stage 3 (What Hurts?), scored by frequency × intensity.
  3. Winning angle from stage 4 (How to Win?), tied to which of the 7 Powers the business will have at maturity.
  4. V1 scope from stage 5 (What’s V1?). Differentiator + Operational features; Delights explicitly held back for V2.
  5. Pricing model from stage 6 (How to Charge?). Tier structure, entry price, the Pricing Position verdict.
  6. Launch plan from stage 8 (How to Launch?). Primary channels and message.

ShipFit also installs project-level shortcuts that re-apply the framework gates (MoSCoW, Van Westendorp, Superhuman PMF Engine) during development, so the validation discipline survives past the planning meeting. The exact mechanics are handled for you; you get the discipline, not the configuration burden.

Install

After running stage 9, unzip the export into your repo root and commit. The next Claude Code session loads everything automatically. No CLI flags, no plugin install.

Why bother

Three weeks into a build, the conviction from your validation work is fading. Your codebase fills with little exceptions: a feature for “that one user,” a pricing-model special case, a launch channel that wasn’t in the plan. Each one looks reasonable in isolation. Together they mean V1 ships solving a different problem than the one you validated.

Loading the validated playbook into Claude Code’s project context is the cheapest insurance against that drift. The validated decisions stay in the model’s view for every prompt, for the life of the project. When the agent is about to wander off scope, the validation context pulls it back.

It’s not magic. It’s just discipline that’s hard to forget because it’s sitting in the repo.

Common mistakes

1. Treating the export as a one-time setup. Re-export after any meaningful pivot. The repo’s validation context should always match your current ShipFit state, not the version you started with.

2. Letting the framework gates collect dust. The first time the gate flags scope creep, the temptation is to dismiss it. The flag means it caught real drift. Investigate.

3. Mixing engineering conventions with validation context in the wrong place. Validation goes in one block, engineering conventions (test framework, file layout, linting) in another. Re-exporting refreshes the validation block; your engineering block stays put.

Further reading

Frequently asked questions

Why use a ShipFit export instead of pasting the playbook into a Claude Code prompt?
Pasting works for one session. A persistent project context loads into every session automatically, so the validation context never decays. Three weeks in, when you ask Claude Code for 'a new admin dashboard,' the agent already has your V1 scope, your buyer's verbatim pain phrasing, and the pricing model loaded. The cost of pasting is that you eventually stop pasting; the cost of a persistent context is zero after install.
Do I need a paid Claude Code plan for this to work?
Yes for production use. Claude Code requires an Anthropic account and the free tier is rate-limited. The ShipFit export works on every Claude Code tier (Pro, Max, Team, Enterprise) and uses no API beyond what your Claude Code session already consumes.
Can I edit the project context after ShipFit generates it?
Yes. The export is a starting point that captures every validated decision; you add engineering conventions (test framework, file layout, linting) on top. Re-running stage 9 in ShipFit refreshes the validation sections without disturbing the engineering parts you added by hand.
What if my validation outputs change later?
Re-run stage 9. Each export refreshes the validation context to match your current ShipFit state. Most teams re-export after a meaningful pivot (new buyer segment, new pricing model). The framework checks are baked into the export so they re-emit automatically.
How is this different from the Cursor export?
Same validated playbook, different tool. Claude Code is terminal-native; Cursor is IDE-native. Stage 9 in ShipFit lets you pick both (or all the supported tools) and emits the right configuration for each. Most founders use Cursor for editing and Claude Code for terminal/automation; ShipFit pre-configures the pair to share the same validation context, so the two tools agree about what's in scope.

Ready to make your next product a success?

9 decisions between your idea and a product worth building.

No credit card required.

Try an example: