prompty
Prompty makes it easy to create, manage, debug, and evaluate LLM prompts for your AI applications. Prompty is an asset class and format for LLM prompts designe
What is prompty?
Prompty makes it easy to create, manage, debug, and evaluate LLM prompts for your AI applications. Prompty is an asset class and format for LLM prompts designed to enhance observability, understandability, and portability for developers.
Overview of prompty
What it is
Prompty is a markdown file format (.prompty) for LLM prompts. It allows developers to write a prompt once and run it from VS Code or any supported language runtime. The format consists of YAML frontmatter (model configuration, inputs, tools) and a markdown body (the prompt with role markers and template syntax).
Key features
- Cross-runtime
.promptyformat shared across Python, TypeScript, Rust, C#, Java, Go, and Swift - Support for OpenAI-compatible endpoints, enabling endpoint switching without changing the prompt file - Agentic turns (v2): tool loop, runtime controls, context and memory, observable agents - VS Code extension with live preview, connections sidebar, chat mode, and trace viewer - Template syntax using Jinja2 or Mustache - Role markers (system:,user:,assistant:) for message boundaries
Use cases
- Creating, managing, debugging, and evaluating LLM prompts for AI applications - Writing a prompt once and executing it across different runtimes and model providers - Building agentic applications with tool calling, memory, and runtime guardrails - Switching between OpenAI, Microsoft Foundry, or Anthropic endpoints without modifying the prompt asset - Observing the full execution pipeline (render, parse, model call, tool call) via trace files
Getting started
Write a .prompty file: prompty --- name: greeting model: id: gpt-4o-mini provider: openai connection: kind: key apiKey: ${env:OPENAI_API_KEY} template: format: kind: jinja2 parser: kind: prompty --- system: You are a friendly assistant. user: Say hello to {{name}}. Run it with Python: bash uv pip install 'prompty[jinja2,openai]' python import prompty result = prompty.invoke('greeting.prompty', inputs={'name': 'Jane'}) print(result) Or with TypeScript: bash npm install @prompty/core @prompty/openai typescript import { invoke } from '@prompty/core'; import '@prompty/openai'; const result = await invoke('greeting.prompty', { name: 'Jane' }); console.log(result); In VS Code, open the .prompty file and press F5.
License and pricing
Licensed under the MIT License. Open source.
Based on the project's public README and repository metadata.