From zero to your AI reaching your machine.
Four steps. About a minute each. No cloud console, no proxy to configure, no networking chapter to read first.
- Step 1 · ~30 seconds
Create your account
Sign in with Google or GitHub. No credit card, no email verification, no cold-start survey.
- Step 2 · ~1 minute
Add your first connector
Pick one from the catalog — filesystem to reach your files, kubectl to reach your cluster, github to reach your repos. Scope what it can touch. That's it.
- Step 3 · ~30 seconds
Install the desktop app
One binary for macOS, Windows, or Linux. It auto-picks up the MCP you configured and opens a per-user encrypted tunnel to us.
- Step 4 · ~1 minute
Connect your AI
Grab the URL, paste it into your AI. It reaches every connector you set up, only while your desktop is running.
New to MCP? It's the shared language AI apps use to reach tools and data outside themselves — like a universal port. Claude, ChatGPT, and others already speak it, which is why they all work here.
Anything that speaks the Model Context Protocol should work — this is a shortlist of apps we know speak it, not a promise we've end-to-end tested every one. If you hit a snag with one of these, or use one we didn't name, tell us.
You just ask, in plain language.
Type it into the AI you paired in step 4 — from wherever you are, whatever's in your hand.
The bits people ask before signing up.
What is MCP?
Why configure the MCP before installing the client?
What if my machine restarts or the app crashes?
What data actually leaves my machine?
What does folder-scoping mean, exactly?
Can I revoke access at any time?
Does it work offline?
Give it four minutes.
Free forever for two MCPs. No card, nothing to uninstall if it isn't for you.
macOS · Windows · Linux — one small binary