Skip to main content
The fastest way to set up Keystroke is giving your coding agent the following prompt:
If you prefer to run the steps yourself, this guide will install the CLI, scaffold a Keystroke codebase, deploy it to a platform project, run the example, then show prompts you can give your coding agent. If you prefer to build with no local setup, you can connect the Keystroke MCP server to ChatGPT, Claude, Cursor, etc and build via MCP instead.

​
Prerequisites

Before you start, you’ll need Node.js ^22.18.0 || >=24.0.0 (Node 23 is not supported).

​
1. Install the CLI

Install the keystroke command globally.

​
2. Connect the CLI

Link the CLI to your Keystroke account.

​
3. Scaffold a new project

Scaffold a new project from the default template.
This creates keystroke.config.ts, a committed tsconfig.json, a src/ directory with an example agent and workflow, and an AGENTS.md guide so your coding agent knows how Keystroke works. Project scripts call the Keystroke CLI for lint, typecheck, build, and test; the template includes TypeScript, Vitest, and Node types for authoring and editor support. Your local codebase is where you build. Your platform project is where Keystroke runs the deployed version for your team.

​
4. Deploy to the platform

List your platform projects:
Then link this directory to your platform project and deploy:
Or seed the link when scaffolding:
Deploy runs lint and typecheck, builds your project, and uploads the dist/ artifact. Your agents, workflows, and triggers now run in the cloud platform. Pass --project <slug> to deploy to a different project without changing keystroke.config.ts. You do not need to run keystroke lint, keystroke typecheck, or keystroke build first — keystroke deploy runs those checks automatically and fails if anything is wrong.

​
5. Build an automation

Keystroke is built for coding agents. Open your project in Cursor, Claude Code, or another agentic IDE and ask it to help you build. Describe the outcome you want rather than choosing a primitive up front. Your coding agent can inspect available integrations and credentials, clarify consequential choices, and recommend a reliable workflow or an agent where runtime judgment is useful. Example prompts: