Prompty: the first step toward automating your AI workflows in VS Code

In an ecosystem where artificial intelligence is now part of every stage of software creation and development, prompt management quickly becomes a strategic issue. With the rise of large language models (LLMs) like GPT, Claude, Mistral, or Llama, developers often juggle dozens of scattered prompts stored in files, scripts, or notes. The result is a loss of consistency, difficulty testing, and the inability to efficiently automate AI workflows. For a broader overview of prompt organization methods and tools, see our article Organize AI Prompts: Complete Guide to Solutions in 2025.

This is where Prompty, an open source initiative from Microsoft, provides an elegant solution. Integrated into Visual Studio Code, this tool transforms prompt management into a clear, versioned, and reproducible process. It uses a standardized format that encapsulates the text, variables, and model configuration in a single .prompty file. As the official documentation on GitHub explains, Prompty was designed to “improve the portability, traceability, and understanding of prompts.”

Beyond a simple text generator, Prompty is positioned as the first step toward AI workflow automation. By structuring prompts, it lays the foundation for integration into orchestration frameworks such as LangChain, Prompt Flow, or Semantic Kernel, where every element of the pipeline becomes interconnected and reproducible.

What is Prompty? A new format for organizing AI prompts

Prompty is both a standardized format and a Visual Studio Code extension designed to structure and execute prompts for language models. A .prompty file contains all the necessary elements of a prompt: the text, variables, model configuration (OpenAI, Ollama, Azure OpenAI, etc.), and input examples.

This format is based on a clear YAML structure, making it readable, traceable, and easy to version-control. As noted on the official Microsoft GitHub page, the goal is to “standardize prompts and their execution in a single artifact to improve observability and portability.”

In practice, Prompty turns a prompt, once just an experimental string of text, into a true software component. It allows AI teams to centralize, reuse, and share their prompts while maintaining consistency. Compatible with various LLM endpoints (OpenAI, Anthropic, Mistral, Ollama, Hugging Face, etc.), it enables testing the same prompt across multiple AI models without changing its structure.

Prompty in VS Code: centralize and test your AI prompts

One of Prompty’s major strengths is its seamless integration with Visual Studio Code, already a familiar environment for most AI developers. Once installed from the official VS Code Marketplace, you can create a new .prompty file in one click using “New Prompty.”

What sets Prompty apart from a simple text editor is its Preview as Markdown feature: it displays a live preview of the rendered prompt with replaced variables, without running any model. This allows you to refine the phrasing, verify tone consistency, and test the structure before sending it to an LLM.

By pressing F5, Prompty can also execute the prompt directly through any configured model: OpenAI, Azure, Mistral, or Ollama for local use. This flexibility allows developers to compare the behavior of the same prompt across different AI engines while maintaining context and parameter traceability.

Finally, Prompty acts as an AI prompt library manager: each file can be stored, versioned, and organized within a dedicated workspace (text generation, analysis, summarization, fact-checking…). For developers and content creators whose prompts are scattered, this centralization provides a significant gain in productivity and clarity.

Why Prompty is the first step toward automated AI workflows

Before automating an artificial intelligence workflow, you must stabilize its building blocks: the prompts. In an AI pipeline, prompts define logic, inputs, and model behavior. Without a clear structure, orchestrating a set of agents or tasks efficiently is impossible.

This is where Prompty plays a central role. It transforms each prompt into a standardized and versioned component, ready for integration into an automated system. As explained in the Prompty documentation, the tool aligns with modern orchestration frameworks like LangChain, LangGraph, Prompt Flow, and Semantic Kernel.

For example, a developer creates a news summary prompt in Prompty, tests it locally, validates the output, then integrates it directly into a Prompt Flow pipeline or a LangChain chain without rewriting anything. The same .prompty file becomes a reusable module within a complete AI pipeline.

This standardization simplifies maintenance and traceability. If a prompt evolves, updating a single file is enough. All connected orchestrators (LangChain, LangGraph, Prompt Flow, Semantic Kernel) will automatically use the new version, reducing the risk of inconsistencies between testing and production environments.

Practical guide: create, test, and automate your prompts with Prompty

1. Prepare your environment

Create a working folder, for example: C:ProjectPrompt-EngineeringVS-Code-Prompty

Open it in VS Code, install the Prompty extension, then create a VS-Code-Prompty.code-workspace file with the following configuration:

{
  "folders": [
    {
      "path": "."
    }
  ],
  "settings": {
    "prompty.modelConfigurations": [
      {
        "name": "mock-local",
        "type": "openai",
        "base_url": "http://localhost:0/v1",
        "api_key": "mock",
        "description": "Mock model for local preview only (no external call)"
      },
      {
        "name": "openai-cloud",
        "type": "openai",
        "base_url": "https://api.openai.com/v1",
        "model": "gpt-4o",
        "api_key": "${env:OPENAI_API_KEY}",
        "description": "Live OpenAI model GPT-4o"
      },
      {
        "name": "azure-openai",
        "type": "azure_openai",
        "api_version": "2024-10-21",
        "azure_endpoint": "${env:AZURE_OPENAI_ENDPOINT}",
        "azure_deployment": "gpt-4o-deployment",
        "api_key": "${env:AZURE_OPENAI_API_KEY}",
        "description": "Azure OpenAI GPT-4o deployment"
      }
    ],
    "prompty.defaultModel": "mock-local"
  }
}

The mock-local configuration prevents any external calls and allows the use of F5 without connecting to a model, or Preview as Markdown to visualize and copy the final prompt. The openai-cloud and azure-openai sections are examples you can modify or remove as needed.

2. Create a prompt template

Create a file called 1-collect-AI-news.prompty:

name: AI News Snapshot
description: Summarize key AI announcements between two dates
sample:
  start_date: 2025-10-01
  end_date: 2025-10-07
system:
You are a research assistant specialized in AI.
Your task is to summarize key AI-related announcements between {{start_date}} and {{end_date}}.

Press F5 to preview the final result with variables replaced.

3. Organize your prompt library

Structure your prompts into folders:

📁 VS-Code-Prompty
 ┣ 📂 collect
 ┃ ┗ 1-collect-AI-news.prompty
 ┣ 📂 verify
 ┃ ┗ 2-verify-AI-news.prompty
 ┣ 📂 write
 ┃ ┗ 3-write-weekly-report.prompty
 ┗ 📂 review
   ┗ 4-review-final-report.prompty

Each file becomes a building block for your future AI workflow: collection, verification, writing, review.

4. Local execution with Ollama

Install Ollama, then configure:

{
  "name": "ollama-local",
  "type": "ollama",
  "base_url": "http://localhost:11434",
  "model": "llama3:8b"
}

You can now run prompts locally without cloud access, ideal for offline testing or privacy-sensitive projects.

5. Integration with LangChain or Prompt Flow

In LangChain, a Prompty file can be loaded like this:

from langchain.prompty import PromptyTemplate

template = PromptyTemplate.from_file("collect/1-collect-AI-news.prompty")
prompt = template.format(start_date="2025-10-01", end_date="2025-10-07")
print(prompt)

In Prompt Flow, simply link the .prompty file as a component of a flow.

Prompty, a bridge toward full AI workflow automation

Prompty serves as the foundation layer for modern AI pipelines. Each .prompty file becomes a reusable, versioned, and traceable block that can be connected to orchestrators such as LangChain, Prompt Flow, or Semantic Kernel.

On GitHub, Microsoft already provides examples of direct integration with Prompt Flow, where Prompty acts as a “template loader.” In Semantic Kernel, a Prompty prompt can be converted into an AI function, ensuring consistency between local development and production deployment.

This modular approach reduces friction between experimentation and automation. Prompty therefore becomes the natural bridge between prompt engineering and industrialized AI workflows.

A fast learning curve for VS Code users

Prompty integrates smoothly with the Visual Studio Code ecosystem. Installation is instant, the syntax is clear, and the Preview as Markdown feature allows for offline work. For users who want to go further, the F5 command enables direct execution on the selected model (mock, Ollama, Azure, or OpenAI).

As noted in the Microsoft documentation, Prompty’s goal is to “accelerate the inner development loop.” For anyone already familiar with VS Code, the learning curve is practically non-existent.

FAQ – Common questions about Prompty and VS Code

Is Prompty free? Yes, it is an open source project from Microsoft available on GitHub.

Do I need an API key to use it? No. The Preview as Markdown feature lets you visualize prompts without executing any model.

Does Prompty work with Ollama? Yes, via a local endpoint (http://localhost:11434/v1) compatible with Llama 3, Mistral, or Gemma.

What’s the difference between Prompty and Prompt Flow? Prompty is designed to design and test prompts, while Prompt Flow orchestrates complete AI workflows.

Prompty, a key building block for structured AI workflows ?

Prompty is not just a text editor for AI. It is a software engineering tool built to structure and secure the work of AI teams. By standardizing prompt creation, testing, and reuse, it makes AI workflows more transparent, reproducible, and interoperable.

For developers already working within Visual Studio Code, Prompty offers a natural gateway to advanced prompt engineering and AI workflow automation.

In other words, Prompty marks the shift from handcrafted prompts to industrial-grade AI workflows, the first concrete step toward efficient, maintainable, and automated intelligence systems.

Also read : How to Build AI Agents Independent from Any LLM Provider