How Keystile works

Keystile sits between your AI assistants and your firm’s apps. Your assistants ask Keystile to do things; Keystile checks each request and only then does it, with the app access you connected.

Your AI tool  ──►  Keystile  ──►  Your apps (GitHub, Gmail, YouTube, your CRM, … 750+ in all)
 (Claude,          checks every
  Cursor, …)       action first

Explore first — no account needed

• Find apps for what you want to do — type a task, see the apps Keystile would use.
• Browse all apps — every app and everything it can do.
• Watch the demos — Keystile stopping an agent, working in GitHub, and for each role.
• Ready to try it with your own apps? Get early access.

The two places you work in

	Address	What you do there
Keystile Automations	https://automations.keystile.com	Connect your apps, build workflows, connect your AI tools
Keystile console	https://demo.keystile.com	Approve held actions, set rules, limits and budgets, see what happened

Note

If your firm runs Keystile on its own servers or in its own cloud, your admin will give you your firm’s addresses. Everything else on these pages is the same.

Three steps

1. Sign in
2. Connect your apps — once, for everyone on your team who should use them
3. Connect your AI tool — Claude, Claude Code, Cursor, ChatGPT and others

Then try the first prompts to see Keystile allow, hold and stop actions.

What happens to each action

Decision	What it means	Example
Allowed	It runs straight away	Read the open GitHub issues
Held	It waits for a person to approve or reject it	Open a client’s account documents
Stopped	It never runs, and the AI tool is told why	Delete a code branch; send a Social Security number
Masked	It runs, but sensitive details are hidden from the AI tool	Account numbers in a document the tool reads

Every decision, with its reason, goes into the signed record.