Connect Peleka to an AI assistant (MCP)

Give Claude, ChatGPT, or another MCP-compatible AI assistant access to your Peleka workspace.

Peleka connects to AI assistants like Claude and ChatGPT through a standard called MCP (Model Context Protocol). Once connected, you can ask the assistant to look things up or take action in your workspace the same way you'd ask a teammate, right inside your normal chat with it: no dashboard, no copy-pasting data back and forth.

What is MCP?

Think of it as a USB port for AI tools. MCP is an open standard that lets an assistant plug into an outside app, like Peleka, and use it the same way it might use a calendar or a spreadsheet. You don't need to know anything about how it works under the hood to use it.

What you can ask it to do

Once connected, you can talk to your assistant in plain language. A few examples:

"Draft a re-engagement campaign for contacts who haven't opened anything in 60 days."
"How did our last broadcast perform compared to the one before it?"
"Build a welcome sequence for new signups: an intro email right away, then a follow-up two days later."
"Pull up [email protected]'s activity, has she opened anything recently?"
"Tag everyone in the VIP segment as priority-support."
"Is our sending domain warmed up enough to email our full list today?"

The assistant is working with the same contacts, broadcasts, automations, segments, tags, and forms you'd see in the dashboard. It can look things up, draft campaigns and automations, tag and enroll contacts, and check analytics and send-readiness. It writes the actual email copy itself when you ask it to draft something; it's not filling in a template.

How to connect

The steps are roughly the same regardless of which assistant you're using:

In your AI tool's settings, find the option to add a custom connector or MCP server. The exact wording and location vary by tool and change over time, so check that tool's own docs if you can't find it.
Enter Peleka's MCP URL: https://api.peleka.io/api/v1/mcp
Your assistant redirects you to Peleka to log in, if you aren't already.
You'll land on a screen asking which workspace to connect and what level of access to grant. Pick a workspace, choose read or write (more on the difference below), and approve.
You're redirected back to your assistant, connected and ready.

There's nothing to copy-paste and no separate account to set up. It's the same kind of "sign in with Peleka" flow you've probably seen with other apps.

Read access vs. write access

Read access lets the assistant look things up: contacts, broadcast performance, segment membership, analytics, and so on. Nothing changes in your workspace.

Write access adds the ability to actually do things: tag a contact, build an automation, draft a broadcast, send one (with your confirmation, see below). If you only want your assistant answering questions, read access is the safer default.

Only workspace admins can grant write access

If you connect at read scope and later ask your assistant to send a campaign or enroll someone in an automation, it'll fail with a permissions error rather than silently doing nothing. That's expected: granting write access is deliberately limited to workspace admins, the same restriction that already applies to creating a write-scoped API key. If you're not an admin and need write access, ask one on your team to connect it, or to bump your role.

What it can't do

A few things are intentionally out of reach, even with write access:

It can't delete anything. There's no path from a chat message to a deleted contact, broadcast, or automation.
It can't create new tags or segments. Existing tags can be added to or removed from a contact; existing segments can only be added, not removed, and only while creating or updating a contact.
It can't touch billing, team members, domain or sending setup, custom fields, or webhooks. Those stay dashboard-only.
It can't import a CSV of contacts.

If you ask for one of these, the assistant will typically tell you it can't do it rather than attempting something else in its place.

Nothing sends without your say-so

Sending or scheduling a broadcast always goes through a two-step confirmation, even with write access granted. The assistant first previews exactly what would go out: subject line, sender, and how many people would receive it. Only then can it actually send. It can't skip straight to sending, and it can't reuse an old confirmation for a broadcast that's changed since you saw it. If you've asked your assistant to draft and send something, expect to see and approve the details before it actually goes out.

Disconnecting

Go to Settings → Connected Apps to see every AI assistant currently connected to your workspace, including its access level and when it was last used. Revoking a connection there takes effect immediately: the assistant loses access right away, and reconnecting means going through the approval steps again.