Promptel vs Prompty.
Two open-source projects that treat prompts as assets rather than strings. Promptel leans into a small DSL with first-class technique blocks. Prompty leans into a YAML-fronted Jinja format with strong VS Code tooling and a wider ecosystem story via Microsoft Promptflow.
| Dimension | Promptel | Prompty |
|---|---|---|
| Primary origin | Skelf Research, MIT, open source | Microsoft, MIT, open source, VS Code-led tooling |
| Authoring surface | .prompt DSL (Chevrotain parser) + equivalent YAML | .prompty file: YAML front-matter + Jinja2 template body |
| Runtime language | Node.js (≥ 18); ships as npm package | Python SDK (prompty); also Microsoft Promptflow integration |
| Providers (first-party) | OpenAI, Anthropic, Groq via official SDKs | OpenAI, Azure OpenAI; community runtimes for others |
| Technique vocabulary | chainOfThought, fewShot, zeroShot, treeOfThoughts, reAct, selfConsistency as AST nodes | Free-form; techniques live in the Jinja body or are layered by the caller |
| Reasoning channels | Harmony Protocol channels surfaced as structured fields | Not a native concept; depends on model + harness |
| Typed params | Typed with defaults and optionality, enforced at parse time | Typed inputs/outputs declared in YAML front-matter, validated by runtime |
| Output schema | Declared in the prompt; JSON mode lowered per provider | Output schema in front-matter; tooling-driven enforcement |
| CLI | promptel: execute, convert .prompt ↔ YAML | prompty: execute prompts from CLI; tighter VS Code integration |
| Editor story | Plain text; syntax-highlighting friendly via JS rules | Official VS Code extension with first-class authoring UX |
| Scope | Small library: parse, execute, convert, that is it | Asset format + ecosystem (Promptflow, observability, eval tooling) |
| When to pick it | Node-first team that wants typed, technique-aware prompts in source control | Azure-heavy or Python-first team already in the Microsoft AI tooling lane |
Pick Promptel when
- You are in Node and want a small, single-purpose library.
- Your prompts use named techniques and you want them reviewable as AST nodes.
- You need to swap between OpenAI, Anthropic, and Groq without rewrites.
- You want both an authoring DSL and a CI-friendly YAML form for the same asset.
Pick Prompty when
- You are Python-first and already in the Azure / Microsoft AI tooling lane.
- You want a polished VS Code authoring experience out of the box.
- You plan to use Promptflow for orchestration and observability.
- You are comfortable expressing techniques in Jinja rather than as named blocks.
Sources: github.com/skelf-research/promptel, github.com/microsoft/prompty. If anything here is out of date, please open an issue.