Scalekit AuthStack
Skills and plugins that let your AI coding agent set up AgentKit and SaaSKit for you.
Auth for AI agents and B2B apps takes weeks: OAuth flows, SSO providers, SCIM provisioning, MCP server auth, and sessions. AuthStack gives your coding agent the steps to do that work in your project, with checkable results at the end of each step.
It works in Claude Code, Cursor, Codex, GitHub Copilot, and any agent that reads Agent Skills.
Choose a kit
| AgentKit | SaaSKit |
|---|
| Use it when | Your agent acts in a user's apps | Your app needs login and enterprise auth |
| What you get | OAuth, a token vault, connections, and tools. 600+ connectors and 35,000+ tools. Every call runs as the user who authorized it. | Login, sessions, SSO (SAML and OIDC), SCIM, RBAC, MCP server auth, and API keys |
| Start with | setup-agentkit | setup-saaskit |
Not sure? Ask your agent "which Scalekit kit should I use?" and setup-scalekit picks one.
Install
Pick one path.
Scalekit CLI. It detects the AI coding tools on your machine and sets up the right plugin:
npx @scalekit-inc/cli setup
To target one tool:
npx @scalekit-inc/cli setup claude
npx @scalekit-inc/cli setup cursor
npx @scalekit-inc/cli setup codex
npx @scalekit-inc/cli setup copilot
Marketplaces.
| Tool | Install |
|---|
| Cursor | Open Customize and install scalekit-agentkit, scalekit-saaskit, or both from the Cursor Marketplace. Plugin docs. |
| Claude Code | /plugin marketplace add scalekit-inc/authstack, then /plugin install agentkit@authstack or /plugin install saaskit@authstack |
| GitHub Copilot | copilot plugin marketplace add scalekit-inc/authstack, then copilot plugin install agentkit@authstack or copilot plugin install saaskit@authstack |
Skills only. Any agent that reads Agent Skills can install the same skills straight from this repo:
npx skills add scalekit-inc/authstack --list
Requirements
- A Scalekit account. A free tier is available.
SCALEKIT_ENVIRONMENT_URL, SCALEKIT_CLIENT_ID, and SCALEKIT_CLIENT_SECRET. The setup skills walk you through them.- Cursor: the
scalekit MCP server at mcp.scalekit.com asks for OAuth on first use.
Your first prompt
After you install a kit, open your project and ask in plain English:
- AgentKit: "Setup AgentKit in this project."
- SaaSKit: "Setup SaaSKit in this project."
Every skill is model-invoked, so your agent picks the right one from what you ask. Each skill ends on a result you can check, such as a working login URL or a connected account.
What's inside
Each kit ships skills. Folder names match skill names.
| Skill | What it does |
|---|
setup-agentkit | Sets up a dashboard connection, env credentials, and a first connector path |
integrate-agentkit | Creates a connected account, authorizes it, then calls one tool |
discover-connectors | Finds connectors, tools, and schemas from the live catalog |
expose-agentkit-mcp | Exposes AgentKit tools over MCP |
integrate-agentkit-host | Connects an always-on host such as OpenClaw or Hermes |
check-agentkit-prod | Runs the go-live checklist |
| Skill | What it does |
|---|
setup-saaskit | Sets up env credentials, a redirect URI, and a first login URL |
implement-saaskit, implement-saaskit-nextjs, implement-saaskit-python | Adds login, callback, session cookies, and logout |
manage-saaskit-sessions | Stores, validates, refreshes, and revokes sessions |
implement-access-control | Adds roles and permissions to routes |
implement-sso | Adds SAML and OIDC single sign-on and the admin portal |
implement-scim | Provisions users and groups from a directory |
add-mcp-oauth | Protects your MCP server with OAuth |
add-api-auth | Adds API key or client-credentials auth |
migrate-to-saaskit | Audits and imports existing auth |
run-dryrun | Tests an auth setup before go-live |
review-scalekit-code | Checks a Scalekit code snippet |
deploy-self-hosted | Deploys self-hosted Scalekit with Helm |
check-saaskit-prod | Runs the go-live checklist |
setup-scalekit installs Scalekit and picks a kit. It installs through the CLI and npx skills add. The Cursor plugins do not include it.
Resources
Contributing
Read AGENTS.md first. It covers the writing bar for skills, versioning, and the check to run before a pull request:
scripts/validate.sh
See CHANGELOG.md for releases.
Security
Report vulnerabilities privately. See SECURITY.md.
Legal
MIT License · Terms of Service · Privacy Policy
