A declarative language for describing applications in natural prose that AI agents compile to source code.

Unlike traditional programming languages that use syntax-tree compilers, prose-lang uses an LLM Agent as its compiler.

You do not run a traditional binary to build your app. Instead, you interact with your AI IDE (like GitHub Copilot, Cursor, or Gemini) equipped with the Prose-Lang Agent Skill to execute the complete software development lifecycle:

1. Write Specs (write .prose): You author .prose files detailing memory, layout, and behaviors.
2. Generate Code (prose generate): The AI Agent reads your .prose specs and compiles them into real, syntax-correct source code.
3. Build the App (prose build): The generated code is built into a runnable application.
4. Test & Validate (prose test): Tests are run to ensure validation of the implementation against your original specs.
5. Document (prose documentation): User-friendly documentation is automatically generated.
6. Package & Publish (prose publish): The finalized application is packaged and prepared for distribution.

# Context
Type: Command Line app (CLI)
Stack: Go

# Memory
NONE

# Behaviors
1. Say 'Hello World'

# Interface
NONE

In the era of AI-assisted engineering and "vibe coding," prose-lang acts as the architectural anchor that keeps your projects scalable and maintainable.

1. Architecture First, Zero Spaghetti: Vibe coding often leads to tangled state and inconsistent abstractions as the AI organically patches code. Prose-lang enforces a strict top-down architecture (Memory, Behaviors, Interface) before generation, ensuring clean domain models and clear boundaries.
2. Infinite Context Memory: LLMs have finite context windows and start hallucinating as codebases grow beyond a few thousand lines. A .prose file acts as heavily compressed memory. You can hand a .prose file to an AI months later, and it will instantly understand your exact business rules without needing to read 100 source files.
3. The Ultimate AI "Reset Button": When a long AI chat session gets confused and starts introducing regressions, you don't have to start over from scratch. You can start a fresh chat, attach your .prose file, and say, "We are taking over a project matching this spec." The AI instantly aligns to the domain architecture perfectly.
4. Language & Vendor Agnosticism: Your core business logic is abstracted entirely away from the syntax. If you decide to port your backend from Python to Go, or swap out React for Vue, the .prose file makes it trivial. The logic survives the rewrite.
5. Instant Auditing & Human Readability: Reviewing a 100-line Markdown specification for business logic flaws, edge cases, or security holes is infinitely faster and more reliable than reviewing 2,000 lines of generated, syntax-heavy source code. It bridges the gap between software engineers and product managers.

Requires uv and Python 3.8+.

The recommended way to install prose-lang is via the pre-built release on GitHub. This CLI tool installs the agent skill instructions into your repository so your LLM knows how to compile prose.

# Install the latest release directly via URL
uv tool install https://github.com/dhilipkumars/prose-lang/releases/download/v0.2.0-alpha/prose_install-0.0.1a0-py3-none-any.whl

Alternatively, to build from source:

git clone https://github.com/dhilipkumars/prose-lang.git
cd prose-lang
uv tool install .

# Install the skill instructions into the current project directory
prose-install --scope local --agent gemini

# Install globally (into ~/)
prose-install --scope global --agent claude

See the examples/ directory for sample .prose files demonstrating various application types, including CLI tools, full-stack web apps, and microservices. Note how the specs live in src/ and output to generated/.

Once the skill is installed, simply chat with your AI and ask it to execute the following workflows:

We welcome contributions! Please see our Contributing Guide and Code of Conduct for details on how to get started, run tests, and submit Pull Requests.

uv tool install . --force --reinstall

MIT

• Spec Kit Integration: Fully test and support seamless integration as an extension for GitHub's Spec Kit.

  # Future flow
  specify init my-project --ai gemini --ai-skills

Note: prose-lang is currently an experimental research project exploring AI-driven compilation. For now avoid in production environments.