Skip to main content
Norbelys is AI-first, and that goes beyond the built-in assistant. You can connect an outside AI agent — Claude, Cursor, an assistant your team built, or anything that speaks the same standard — and it gets the whole Norbelys platform as tools it can use: import leads, build segments, create and launch campaigns, verify emails, read analytics, and more. This works through a connector (an MCP server) that Norbelys hosts for you. You point your agent at one address, sign in, and it can operate your workspace.
This is different from Norbe, the assistant that lives inside Norbelys. Norbe is always there and needs no setup. This page is about giving an AI tool you use elsewhere access to your Norbelys data. See Norbe.

​
What an agent can do

Once connected, an agent can do essentially anything you can do in the app, for example:
  • Confirm which workspace it’s in and get a one-look briefing of your campaigns, mailboxes, and deliverability.
  • Import a lead list and verify the addresses.
  • Build a segment (“verified SaaS founders we haven’t contacted”).
  • Create a campaign, write the sequence, attach mailboxes, and enroll the segment.
  • Launch it — and then monitor replies, opens, and deliverability.

​
How to connect

You have two ways to give an agent access, depending on the tool:

Sign-in flow (easiest)

Tools that support remote connectors (like Claude’s custom connectors) just need the connector address. You sign in with your Norbelys account and pick the workspace during consent — no key to copy.

API key

Headless tools, scripts, and coding assistants use an API key as a bearer token. The same key works for the API and for agents.
The exact connect steps for popular tools (Claude, Cursor, and generic clients) are in the developer guide:

MCP server setup

The connector address, how to authenticate, and copy-paste config for common AI tools.

​
Safety — an agent is still on a leash

Connecting an AI agent never removes your safety rails. The same protections that apply in the app apply to an agent:
  • Sending is treated as dangerous. Actions that put real email in flight — launching a campaign, resuming one, or sending a message (even a test) — are flagged so a good agent confirms with you first.
  • Your workspace is the boundary. An agent can only ever act on the workspace its credential belongs to — there’s no way for it to reach into another.
  • Every platform rule still holds — suppression lists, one-click unsubscribe, sending windows, and daily limits all apply no matter who or what is driving.
A good pattern: ask your agent to plan the work and show you the steps, review anything that sends, and let it run the read-only and setup steps on its own. Keep the launch step for your own approval.

​
Why this matters

Because the app, Norbe, your code, and outside agents all go through the exact same API, there’s no “second-class” way to use Norbelys. Whatever the best tool for the moment is — clicking, asking Norbe, a script, or your favorite AI assistant — it can do the whole job. See API keys & automation.