A runtime that executes AI prompt pipelines defined in a single markdown file. The markdown is the program, the model is the CPU.
Get the Workshop, the desktop app for writing and running prompts:
- Windows: PromptForge-setup.exe
- macOS (Apple Silicon): PromptForge-arm64.dmg
- macOS (Intel): PromptForge-x64.dmg
- Linux: PromptForge.AppImage or PromptForge.deb
These links always point at the latest tested release. Running a headless gateway instead? Grab a nightly build or build from source below. Embedding the engine in your own program? The guide covers every audience.
The Workshop is the desktop application. It edits prompts as visible stacks of blocks, runs them against your gateway, and records every run, edit, and decision in an append-only event log. Voice input is built in, and the app updates itself from the release channel.
The gateway is the one process that holds your credentials. It serves an OpenAI-compatible API, routes chat completions to frontier APIs or to local models on your own hardware, and keeps vendor keys off every other process. One configuration file defines the model catalog, the concurrency pools, and the search tool.
The prompt language is the programming surface. A prompt is a markdown document: YAML frontmatter for metadata, embedded Lua for logic, prose blocks for model instructions. The engine executes it with deterministic control flow, isolated sections, and fan-out concurrency set in configuration. The same language is available as a Rust library, so your own programs can run prompts in-process.
---
name: greet
description: Greet the named input using a Lua-computed value
promptforge: 1
---
# Greet
```lua
models.default("writer", "A model suited for careful analysis, coding, and general assistance")
```
## Main
```lua
var.greeting = "Hello, " .. args .. "!"
```
Repeat exactly, with no extra words: {{ var.greeting }}Prose goes to the model. Lua sets up the turn. The response is the run's result.
Every build needs Rust 1.89 or later and Node.js 22. The two web UIs are bundled with esbuild during the Cargo build, so run npm ci once in each ui/ folder after cloning:
git clone git@github.com:cppalliance/promptforge.git
cd promptforge
npm ci --prefix crates/workshop-server/ui
npm ci --prefix crates/gateway-config-ui/uicargo build builds the gateway, the default workspace member. cargo build -p workshop builds the desktop app. Platform notes:
- Ubuntu 22.04:
sudo apt install build-essential pkg-config cmake clang libclang-dev; the desktop app also needslibwebkit2gtk-4.1-dev libxdo-dev libssl-dev libayatana-appindicator3-dev librsvg2-dev, and builds withcargo build -p workshop --no-default-features. - macOS:
xcode-select --installandbrew install cmake node, thencargo build -p workshop --no-default-features. - Windows: install Visual Studio with the "Desktop development with C++" workload and Node.js 22, then
cargo build -p workshop.
The first build downloads the tool picker's embedding model (~130MB from Hugging Face, pinned and checksummed). Later builds reuse the cache.
Parse a promptforge markdown file, bind the tools and models it needs, then execute each H2 section in order. Section Lua prepares state; prose becomes a model turn (with a tool loop when tools are in scope); results land in the store or become the run output.
flowchart LR
MD[Markdown prompt] --> Parse[Parse and bind]
Parse --> Sec[H2 sections]
Sec --> Lua[Lua blocks]
Lua --> Model[Model turn]
Model --> Tools[Tools via gateway or local]
Model --> Store[Store artifacts]
Store --> Out[Run result]
- PromptForge Guide - four documentation sets: the Workshop, the gateway, the prompt language, and agent programs
Build the guide locally with mdbook build guide.
Rust 1.89 or later.
Build, format, and test before you open a PR. CI runs cargo fmt --check, clippy -D warnings, and cargo test --workspace.
Distributed under the Boost Software License 1.0.