Skip to main content
PDD (Prompt-Driven Development) inverts the traditional development model: instead of writing code directly, you write prompt files that describe what you want, then let AI generate and maintain the code for you. Prompts are versioned, composable, and regenerable — your code becomes a disposable artifact.

Install PDD

Get up and running in under 5 minutes with uv or pip

Quickstart

Run your first prompt-to-code workflow with the Hello example

Core Concepts

Understand the PDD philosophy and the mold paradigm

Command Reference

Every command, flag, and option documented

​
What PDD does

PDD provides a complete workflow from a natural-language prompt to production-ready, tested code:
1

Write a prompt file

Describe what you want in a .prompt file using natural language. Prompts are modular and composable — one prompt per module.
2

Run pdd sync

The sync command runs the full cycle automatically: resolves dependencies, generates code, creates runnable examples, verifies correctness, generates tests, fixes bugs, and back-propagates learnings to your prompt.
3

Iterate with prompts, not code

When requirements change, update the prompt and regenerate. PDD’s fingerprint-based change detection only regenerates what needs updating.
4

Implement GitHub issues automatically

Use pdd change <issue-url> or pdd bug <issue-url> to let PDD implement or fix GitHub issues end-to-end using an agentic multi-step workflow.

​
Choose your workflow

Web Interface

Launch pdd connect for a visual browser-based UI — the easiest way to get started

Issue-Driven CLI

Point PDD at a GitHub issue and let it implement the feature or fix automatically

Prompt Workflow

Write prompt files and use pdd sync for full manual control over the generation cycle

​
Why PDD?

Traditional development treats code as the source of truth. PDD treats prompts as the source of truth — code is a generated artifact that can be regenerated at any time.
  • Maintenance cost drops — 80–90% of software cost is post-creation maintenance. Regenerating from prompts preserves conceptual integrity and eliminates accumulated patches.
  • Intent is preserved — Prompts capture the “why” behind code. The design rationale is explicit, not buried in comments.
  • Batch leverage — Modern LLMs make full-module regeneration practical and cost-effective. PDD’s cost tracking keeps spend predictable.
  • Works with any LLM — Claude, GPT-4, Gemini, and dozens more via LiteLLM. Run in cloud mode (no API keys needed) or local mode with your own keys.

Build docs developers (and LLMs) love

Get started for freeTalk to us