microsoft/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.

What it solves

Prompty provides a standardized, portable markdown file format (.prompty) for LLM prompts. It decouples the prompt engineering process from the application code, allowing developers to write prompts once and execute them across different language runtimes (Python, TypeScript, Rust, C#, Java, Go, and Swift) and LLM providers (OpenAI, Microsoft Foundry, Anthropic) without changing the prompt body.

How it works

Prompty uses a two-part file structure: a YAML frontmatter for configuration (model ID, provider, connection details, inputs, and tools) and a markdown body for the actual prompt text using role markers (e.g., system:, user:) and template syntax (Jinja2 or Mustache).

To execute a prompt, a Prompty runtime loads the file, renders the template with provided inputs, parses the roles, and sends the request to the LLM provider. For more complex interactions, the turn() function enables "agentic turns," allowing the model to loop through tool calls and manage memory/context budgets before returning a final answer.

Who it’s for

Developers building LLM-powered applications who want a consistent way to manage, version, and test prompts across multiple programming languages and different model providers.

Highlights

  • Cross-Runtime Support: Official packages for Python, TypeScript, and several other languages including Rust, C#, Java, Go, and Swift.

  • VS Code Extension: Includes a live preview, a connections sidebar for managing endpoints, a chat mode for interactive testing, and a trace viewer for debugging the full pipeline.

  • Agentic Capabilities: Supports tool-calling loops, runtime controls (guardrails, retries), and memory-aware turns for portable agent behavior.

  • Portable Assets: Allows switching LLM endpoints and keys via environment variables or runtime settings without modifying the prompt file itself.

  • Citations and Tracing: Every execution generates a .tracy trace file to inspect render, parse, and execution timing and payloads.

Related

  • Project
    google/dotprompt

    Dotprompt is a cross‑language framework that defines an executable `.prompt` file format for generative AI. Each file bundles the prompt text, model configuration, input schema, and expected output schema, enabling validation, type‑safe execution, and reuse across JavaScript/TypeScript, Python, Go, Rust, and Java. The project ships core libraries, IDE extensions, web editors, a tree‑sitter grammar, and a CLI tool, all backed by CI testing. It is ideal for developers and prompt engineers who want structured, version‑controlled prompts that can be run directly from code or the command line.

  • Project
    chr15m/runprompt

    A command-line tool to execute .prompt files, allowing users to treat LLM prompts as versionable artifacts and automate AI workflows via the shell.

  • Project
    mufeedvh/code2prompt

    A context engineering tool that converts entire codebases into structured LLM prompts, automating the process of gathering and formatting code for AI models.

  • Project
    langgptai/LangGPT

    LangGPT is a structured, reusable prompt‑design framework that treats prompts like a tiny programming language. It defines a clear template (role, profile, goal, skills, rules, workflow, etc.), supports variables, commands, and conditional logic, and works across major LLMs. The project provides example libraries, ready‑made ChatGPT/Claude assistants, a Claude Code skill, and an ecosystem of related tools (PromptVer, PromptShow, Minstrel) plus model‑specific prompt collections.

  • Project
    MagnivOrg/prompt-layer-library

    A Python SDK for PromptLayer that enables prompt versioning, tracing, and monitoring for LLM applications and agents across multiple AI providers.