Documentation

Getting started

Learn what SubKit is, get an engine running and build your first workflow, in this order.

1. 01What is SubKit?

   What SubKit is and how governed, recorded workflows help you.

2. 02Use the Cloud or self-host

   Use SubKit Cloud

   Use SubKit as a hosted service without running your own engine.

   Run the engine

   Start the engine with Docker Compose on your own server.

3. 03First-run setup

   Set up providers, admin access and the Cockpit after the first start.

4. 04Your first workflow

   Build, approve and run your first workflow from the Cockpit chat.

Working with SubKit

Reach SubKit from your terminal, an AI client, a coding agent, the browser, the HTTP API or your own app.

• Use the sk CLI on your machine

  Install the sk CLI from your engine and work with it in the terminal.

• Connect AI clients via MCP

  Connect an AI client via MCP so it can find and run your workflows.

• Use SubKit from a coding agent

  Give a coding agent SubKit tools and shared-memory briefings for your repository.

• Collect web pages with the Web Clipper

  Save web pages and selections to your Inbox with the browser extension.

• Collect ideas and turn them into knowledge or skills

  Collect pages and ideas and review which become knowledge or skills.

• Use the HTTP API

  Call SubKit from your own program or script over HTTP.

• Onboard an app

  Connect an app that signs users in at an identity provider to SubKit.

• Reply by email to a signed request

  Let a workflow reply by email to the sender of a signed request.

• Cockpit as an app

  Install the Cockpit of each workspace as its own app with the workspace's name, icon and colors.

Operating SubKit

Run your own engine: branding, sessions, identities, permissions, credentials and sandbox limits.

• Cockpit branding

  Change the Cockpit's name, logos and colors for your workspace.

• Installed Cockpit sessions

  How an installed Cockpit gets and keeps its workspace session.

• Engine operator

  Which workspace operates the engine and what only it may do.

• Execution identity and permission

  Who runs an execution and which permissions apply to it.

• Product identity binding

  Bind your product's signed-in users to a minimal SubKit identity.

• Product bearer tokens and caller budgets

  Let an app call SubKit with its own sign-in and limit each caller's usage.

• Supabase-backed product workflows

  Run product workflows against your Supabase project with the supabase connector.

• External MCP OAuth credentials

  Connect external MCP servers that require OAuth client credentials.

• Direct signed-JWT connector credentials

  Authenticate connectors to APIs that accept a signed JWT directly.

• Sandbox data tools

  Which sk helpers JavaScript transforms can use and where their limits are.

Reference

Look up commands, settings, connectors and the license.

• sk CLI reference

  Every sk command and argument — and the surface authoring runs through: creating or changing a workflow, connector, knowledge source or skill is a definition edit, and definition edits belong here, where the change is a file you can read and a command you can re-run.

• SubKit runtime settings

  Every value below lives in canonical runtime configuration and is editable via the Cockpit, HTTP/MCP settings APIs, sk settings configure, or sk settings set.

• Connector & recipe catalog

  What ships in the box: 61 connector templates across 16 categories, carrying 248 recipes — pre-authored, runnable workflows you clone instead of starting from a blank page.

• License

  Under which license you may use and share SubKit.