Welcome to Prompty

Prompty is a file format (.prompty) for LLM prompts. Write your prompt once — model config, inputs, tools, and template in a single markdown file — then run it from Python, TypeScript, C#, Rust, Java, Go, or Swift. Treat prompts as code: version them in git, test them in CI, share them across teams, and deploy with confidence.

Choose Your Path

🚀 I'm New to PromptyStart from scratch — install, write your first prompt, and run it in under 5 minutes.

🔄 I'm Migrating from v1What changed, what's new, and how to update your existing .prompty files.

⚡ I Know the BasicsJump to how-to guides, tool calling, streaming, structured output, and more.

🔎 I Need a ModelList available models and deployments before setting model.id.

Quick Example

---
name: greeting
model:
  id: gpt-4o-mini
  provider: openai
  connection:
    kind: key
    apiKey: ${env:OPENAI_API_KEY}
inputs:
  - name: userName
    kind: string
    default: World
---
system:
You are a friendly assistant.

user:
Say hello to {{userName}}.

Python

import prompty
result = prompty.invoke("greeting.prompty", inputs={"userName": "Jane"})

TypeScript

import { invoke } from "@prompty/core";
const result = await invoke("greeting.prompty", { userName: "Jane" });

C#

using Prompty.Core;
var result = await Pipeline.InvokeAsync("greeting.prompty", new() { ["userName"] = "Jane" });

Rust

use serde_json::json;

prompty::register_defaults();
prompty_openai::register();

let result = prompty::invoke_from_path(
    "greeting.prompty",
    Some(&json!({ "userName": "Jane" })),
).await?;

What Can You Build?

• Chat assistants — Getting Started
• RAG pipelines — How-To: Embeddings
• Tool-calling agents — How-To: Agent Tool Calling
• Structured data extraction — How-To: Structured Output
• Image generation — How-To: Image Generation

Key Features

• Multi-runtime parity — Python, TypeScript, C#, Rust, Java, Go, and Swift
• Pipeline architecture — render → parse → execute → process, each stage swappable
• Built-in tracing — console, JSON file, and OpenTelemetry backends
• Provider support — OpenAI, Azure / Foundry, and Anthropic
• Model discovery — list available models and deployments where providers expose them
• VS Code extension — syntax highlighting, autocomplete, live preview, and trace viewer

Learn More

📖 Core ConceptsFile format, pipeline, connections, tools, and tracing.

📋 Schema ReferenceEvery frontmatter property documented.

🧪 Live Provider TestsRun endpoint-backed checks before publishing packages.

🐍 Python RuntimeInstallation, API reference, and provider guides.

📘 TypeScript RuntimeInstallation, API reference, and provider guides.

Contributing

Prompty is open source. We welcome contributions to the runtimes, tooling, and documentation. See the Contributing guide to get started.