prompty.ai

What It Is

Prompty is an open file format and runtime for LLM prompt engineering. It combines YAML frontmatter with a Markdown prompt body into a single .prompty file, used to define model configuration, connection details, input schemas, tools, and templates. Compared with scattering prompts throughout code, it is more like a portable prompt asset format that can be executed in Python, TypeScript, and VS Code.

Core Capabilities

Functionally, Prompty covers the full workflow of prompt authoring, rendering, parsing, execution, and result handling. Templates support Jinja2 or Mustache; the execution pipeline consists of four stages—Render, Parse, Execute, and Process—and each stage can be replaced and traced. The VS Code extension is its main workspace, offering syntax highlighting, autocomplete, connection management, live preview, interactive chat, one-click F5 execution, and a built-in trace viewer.

Languages, Integrations, and Observability

Prompty explicitly supports Python, TypeScript, and VS Code. Its connection layer mentions OpenAI, Anthropic, and Microsoft Foundry, and it also allows custom executors. Its tooling support is fairly complete, including local functions, MCP server, OpenAPI tools, automatic tool-calling loops, streaming output, structured output, and the OpenAI Responses API. Observability is a highlight: each run can record messages, tokens, latency, and the raw API response, with support for console output, JSON files, and OpenTelemetry backends. You can also use the @trace decorator to trace your own functions.

Open Source, Pricing, and Documentation

The main content clearly states that Prompty is MIT licensed and provides a GitHub entry point, positioning it as an open standard. No commercial plans are disclosed for pricing, so the only clear conclusion is that the core project is open-source licensed; whether a hosted edition or enterprise services exist is unknown. The site provides entry points such as Get started and Core Concepts, with a clear information structure, but the crawled content does not show full API details, deployment guidance, version compatibility, or support SLA information.

Pros, Cons, and Who It’s For

Its strengths are a unified format, cross-language reusability, built-in debugging and tracing, and integrations with MCP, OpenAPI, and major model providers. Its weaknesses are the lack of information around commercialization, team collaboration, permission management, and self-hosted service models. It is suitable for LLM application developers, prompt engineers, and teams that need to manage prompts in an engineering-oriented way, especially those using Python/TypeScript stacks.

Access from China

The main content does not provide information about access, payment, or mirrors for mainland China, so its availability should be marked as unknown. If network access or model API access is restricted, alternatives or complementary tools such as LangChain/LangSmith, Promptfoo, Dify, Flowise, and Helicone may be worth evaluating.