Prompty

Prompty is an open format for LLM prompts. A prompt, its model configuration, input schema, tools, template and agentic controls live together in a single markdown file, with settings in YAML frontmatter. The same file can then be executed across Python, TypeScript, C# and Rust runtimes without rewriting. The tooling centers on a VS Code extension that works as a prompt engineering workbench. It provides syntax highlighting and autocomplete, connection management for OpenAI, Anthropic and Microsoft Foundry, live preview with an interactive chat mode, and one click execution with inline results. Every run produces a trace covering the messages sent, tokens used, latency and the raw API response, with console, JSON file and OpenTelemetry backends, and a trace viewer inside VS Code. Developers can also decorate their own functions with a trace decorator for end to end visibility. Prompts pass through a four stage pipeline of render, parse, execute and process. Templates can use Jinja2 or Mustache, and tools can be defined as functions, MCP or OpenAPI declarations. Version 2 adds an agentic turn runtime with guardrails, steering, cancellation, retries, context budgets, compaction and memory aware conversations. The format is open and extensible, aimed at developers who want portable, observable prompts. Providers are pluggable, and custom executors can be built. The site publishes no pricing and presents the project through its documentation and GitHub repository.