Turns a Git repository or local directory into a single prompt-friendly text digest for feeding code to LLMs.
promptdriven/pdd
A prompt-native development system where .prompt files are the source and Python, TypeScript, Go and other code is generated from them.
- Stars
- 913
- Forks
- 82
- Open issues
- 493
- Last commit
- 2 days ago
What is it?
PDD (Prompt-Driven Development) treats `.prompt` files as the human-authored source of a project, with traditional language code produced from them. Developers write intent, constraints, examples and tests in prompts, then generate implementation code that stays real and reviewable but is no longer the primary source of truth.
The CLI, installed as `pdd-cli`, provides a `pdd sync` loop that runs dependency insertion, generation, examples, crash handling, verification, testing and fixing for each module. Agentic commands such as `pdd change`, `pdd bug` and `pdd fix` take GitHub issue URLs and run multi-step workflows against Claude Code, Gemini/Antigravity, Codex or OpenCode. A `pdd connect` command opens a local web interface on port 9876.
PDD uses LiteLLM for model access in local mode, with a CSV model catalog and a setup wizard that detects API keys and agentic CLI tools. Cloud mode is the default and uses GitHub sign-in.
Key features
- 01Generates Python, TypeScript, Go and other language code from `.prompt` source files.
- 02Runs a `pdd sync` loop covering dependency insertion, generation, examples, crash handling, verification and tests.
- 03Implements GitHub issues through `pdd change`, `pdd bug` and `pdd fix` commands.
- 04Provides a local web interface launched with `pdd connect`.
- 05Supports multiple agentic CLIs and LLM providers through LiteLLM in local mode.
Use cases
- Maintaining a codebase where prompts describe behavior and code is regenerated when intent changes.
- Turning GitHub feature requests or bug reports into tested changes through the agentic commands.
- Generating the same module in several target languages from one prompt source.
Quick start
- 1
Install uv (run the official install script from the README)
- 2
Install PDD with uv
uv tool install pdd-cli - 3
Run the setup wizard to detect tools and API keys
pdd setup - 4
Launch the web interface
pdd connect
Things to know
- Requires Python 3.12 or higher.
- Needs an LLM API key or a logged-in agentic CLI such as Claude Code, Codex or Gemini CLI to run most workflows.
- The issue-driven CLI workflows also require the GitHub CLI (gh).
In collections
Topics
FAQ
- What is pdd?
- A prompt-native development system where .prompt files are the source and Python, TypeScript, Go and other code is generated from them. PDD (Prompt-Driven Development) treats `.prompt` files as the human-authored source of a project, with traditional language code produced from them. Developers write intent, constraints, examples and tests in prompts, then generate implementation code that stays real and reviewable but is no longer the primary source of truth.
- Is pdd free to use? What license does it have?
- pdd is open source under the MIT license. Check the license terms before using it in commercial products.
- Is pdd still maintained?
- Actively maintained. The last commit to pdd was on Oct 23, 2026.
- What can I use pdd for?
- Maintaining a codebase where prompts describe behavior and code is regenerated when intent changes. Turning GitHub feature requests or bug reports into tested changes through the agentic commands. Generating the same module in several target languages from one prompt source.
- How do I install pdd?
- Install uv (run the official install script from the README) Install PDD with uv: uv tool install pdd-cli Run the setup wizard to detect tools and API keys: pdd setup Launch the web interface: pdd connect
- Where can I get pdd?
- The source code and installation instructions are on GitHub: https://github.com/promptdriven/pdd
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An open source toolkit that gives AI coding agents structured spec-driven processes, templates, and documented outcomes for building features, fixing bugs, or assessing ideas.
A CLI that one-command sets up Claude Code and Codex with bilingual prompts, workflows, API provider presets and MCP configuration.
Turns any GitHub repository into an interactive architecture diagram or a one-minute narrated explainer video, with an MCP server for AI agents.
An agent framework for OpenCode that adds plan-first, approval-gated AI coding workflows using your team's own coding patterns.
An AI prompt optimizer that refines prompts, tests and compares their outputs, and saves them as reusable assets, available as web, desktop, Chrome extension, and Docker.