Promptbook Engine

Your AI.
On your machine.

Hand over an ongoing responsibility, set the rules and let the engine keep working. Agents, tasks and results live in a readable folder under your control.

Explore the source code

Start on your machine.

Choose an installation method. The setup guide prepares your project, AI connection and first workflow.

A few commands. Your own engine.

For macOS, Linux and Windows. Install Node.js 22.13+ in the 22.x release line or version 24+, npm 10+ and Git first.

Terminal

npm install --global ptbk
ptbk --version

mkdir my-agenda
cd my-agenda
ptbk init

From an empty folder to your first workflow.

Copy these commands into Bash or Zsh. They use npm; Node.js and Git must already be installed.

Bash / Zsh

node --version
npm --version
git --version

npm install --global ptbk
mkdir my-agenda && cd my-agenda
ptbk init

Guided installation for Windows.

Choose the Windows x64 EXE on the releases page. The installer adds missing Node.js, npm and Git, installs the engine and offers to create your first project.

Open Assets for your chosen version and download:

Promptbook-Setup-<version>-windows-x64.exeChoose a Windows EXESome releases are marked as prereleases. Choose a version using its release notes.

What happens on first launch?

In ptbk init, describe your ongoing responsibility and follow setup. Choose a supported AI connection and browser; the guide prepares the project and starts work in the background.

Your chosen AI account, its costs and any server are separate. The selected model processes the inputs sent to it.

Keep an eye on progress.

Run these commands one at a time. The dashboard shows progress; close it or use another terminal before opening chat. Closing the dashboard does not stop the engine. Use ptbk stop to shut down gracefully and ptbk start to resume.

ptbk

ptbk dashboard
ptbk chat
ptbk stop
ptbk start

Files you can understand.

Agent definitions, tasks and conversations are readable text. Edit them, version them and take them with you.

The work history is yours.

Git keeps changes to project files. Traces add the available records of how work was carried out.

Your choice of AI.

Use a supported connection to Codex, Claude Code or other harnesses. The model choice stays separate from your agents and context.

README and technical documentation How Agent–Project–Task works

The same engine. Comfortable controls.

Promptbook brings your own engine together with an app and support. Get in touch if you want the complete solution or help setting it up.

Explore the complete Promptbook Deploying for clients? Explore the partnership

Engine licence

Business Source License 1.1 (BUSL-1.1). Find the terms for your version in the repository. LICENSE.md

Like this approach? Give us a star on GitHub.