How to export an implementation plan for AI coding

What it is

Once your project spec is ready, the next question is: where does the AI coding tool start? A PRD describes what you're building — an implementation plan maps out how to build it. It gives your AI coding tool a numbered sequence of steps, files and tables to create, and acceptance checks so it can make real progress without you having to figure out the right order yourself.

The implementation plan is generated directly from your Draftlytic project. It uses your actual spec verbatim — the tech stack, data model, features, constraints — so the AI never has to invent substitute tech or guess at your intent. Only must-have and nice-to-have features are planned; anything you've marked completed becomes an "already built" signal, and future features are listed as out of scope so the AI doesn't accidentally build ahead.

This is a paid feature available on Starter, Plus, and Pro. It costs 40 credits for a standard plan, or 50 credits with the "Include full project context" toggle on (which adds scaffolding steps for project setup, data model, auth, and routing — useful when you're building from scratch).

Step by step

1. 1

   Open your project

   Head to your project detail page in Draftlytic. Make sure your features are set up and have the right priority labels — must-have and nice-to-have features will be planned, completed features are treated as already built, and future features are deferred.

2. 2

   Open the Export implementation plan dialog

   Click the Export button in the project header area and select "Implementation plan" from the menu. A dialog opens showing you exactly which features will be included, grouped by must-have and nice-to-have.

3. 3

   Pick your target AI coding tool

   Use the Target tool dropdown to tell Draftlytic which tool you're building with — Cursor, Windsurf, Lovable, Bolt.new, Claude Code, Replit, GitHub Copilot, and many more are available. The plan's step detail and granularity are tuned to match how that tool works. If you've already saved a preferred tool in your profile, it pre-fills here and you can override it just for this export.

4. 4

   Choose a version scope (if you use milestones)

   If your project has feature milestones, a Version dropdown lets you narrow the plan to a single slice of the roadmap instead of everything at once. Filtering to one milestone keeps the plan focused and avoids overwhelming your AI tool with too many steps.

5. 5

   Decide whether to include full project context

   The "Include full project context" toggle is on by default. Leave it on when you want the plan to start from scratch — it adds scaffolding steps for initial project setup, data model, auth, and routing, and costs 50 credits total. Turn it off for additive mode, where the plan assumes your existing app is already built and only covers the incremental changes for the selected scope. Additive mode costs 40 credits.

6. 6

   Export and download the plan

   Click "Export plan" — the credit cost is shown right on the button so there are no surprises. Draftlytic generates the plan and downloads it as a Markdown file named after your project. If the AI call fails for any reason, your credits are automatically refunded.

7. 7

   Paste it into your AI coding tool

   Open the downloaded Markdown file and paste it as your kickoff prompt. The plan is self-contained — it includes your design, tech stack, and business context alongside the build steps, so you don't need to paste the PRD separately. Your AI coding tool now has a clear, ordered brief it can follow from step one.

Tips

• Mark features as 'completed' before exporting so the plan doesn't rebuild things you've already shipped — completed features become an 'already built' signal the AI respects.
• If you have milestones, filter to one version at a time. Plans with more than 15 features tend to overwhelm AI coding tools — keep each plan tight and focused.
• The plan uses your spec verbatim. The more complete your tech stack and data model are in Draftlytic, the less the AI has to guess when it starts coding.
• Use the target tool selector even if you're not sure yet — picking the closest match produces better step granularity than leaving it on the generic default.
• Credits are auto-refunded if the AI fails, so you won't lose them to a network blip or a timeout. Just try again.
• The PRD and the implementation plan are meant to be used together: paste the PRD for spec alignment, then use the implementation plan as your coding kickoff prompt.

FAQ