You already have ways to talk to your assistant — the app window, Telegram. Those are channels. Connectors are the other half of the picture: they're the services your assistant can reach into on your behalf, where your actual work and data live.
Once you connect a service, you can ask your assistant about it in plain language. "What are my open GitHub issues?" "Update the launch notes in Notion." Your assistant looks it up — or makes the change — using your own access, and answers you in whatever channel you happen to be chatting in.
Connectors vs. channels vs. AI provider
It's easy to mix these up at first. Here's the quick map:
| Piece | What it is | Examples |
|---|---|---|
| Channels | The ways you reach your assistant | Telegram, the app window |
| Connectors | The services your assistant can reach | GitHub, Notion, Stripe, … |
| AI provider | The brain doing the thinking | Claude, OpenAI, … |
Channels are inbound — they carry your messages to the assistant. Connectors are outbound — they let the assistant go fetch and do things in the wider world. And the AI provider is the intelligence tying it all together.
The catalog
HolaClaw launches with eight connectors. Two examples to give you the flavor:
Issues, pull requests, and repository search.
- Authentication
- Sign in with GitHub
- Best for
- Asking about issues and pull requests
Search, read, and update your pages and databases.
- Authentication
- Sign in with Notion
- Best for
- Looking things up and updating docs
Every connector has its own short setup guide with service-specific prompts to try:
- GitHub — issues, pull requests, and repository search.
- Linear — work with issues, projects, and cycles from your Linear workspace.
- Notion — search, read, and update your pages and databases.
- Sentry — inspect issues, traces, and releases when something breaks.
- Stripe — query your customers, payments, and invoices.
- Supabase — manage your projects, tables, and data.
- PostHog — pull product analytics, insights, and feature flags.
- Buffer — draft and schedule social posts.
How signing in works
Every connector uses the service's own sign-in page — exactly like the "Sign in with Google" buttons you see all over the web. You're never typing a password into HolaClaw. Instead:
- You click to connect a service, and your browser opens that service's own page.
- You sign in there (if you aren't already) and approve the access.
- The service hands back an access token — a long secret code that stands in for "this assistant is allowed to act as me here," without ever revealing your password.
That token is stored inside the assistant's private virtual machine on your Mac — the same sealed space that holds your AI provider key and channel config. HolaClaw never sees your password, and our servers never hold your tokens. The connection lives on your machine, between your assistant and the service.
(Under the hood, this all runs on MCP — the Model Context Protocol, an open standard for connecting AI assistants to outside services. You don't need to think about it; it's just the plumbing.)
Per-assistant by design
Connectors belong to a single assistant, not to your whole account. Your work assistant might have GitHub, Linear, and Sentry; your journaling companion has none of them — and connecting GitHub to one does nothing for the other.
That's deliberate. Each assistant gets exactly the reach you give it and nothing more, so a casual companion never gains access to your codebase or your Stripe dashboard just because another assistant has it.
Heads-up — connectors are new. They're rolling out gradually. If you don't see the Connectors tab in your assistant yet, update to the latest version of HolaClaw and check again.
Next steps
Ready to wire one up? Walk through it in Connect a service — one short flow that works the same for every connector in the catalog.
Questions, or want to know which connector is coming next? The HolaClaw Discord is a friendly place to ask.