What are the Landbot MCP and the Landbot Plugin?

Desirée M Updated by Desirée M

AI assistants and AI coding tools, such as Claude, Codex and Cursor, can build, publish and style a Landbot agent for you. You describe the conversation in everyday words, and the AI tool creates the agent in your Landbot account and publishes it when you say so.

The result is a normal Landbot agent. You can still open it in the builder, read its conversations in the inbox and use your integrations and channels as usual.

What can you build with the Landbot MCP and the Landbot Plugin?

The Landbot MCP and the Landbot Plugin build Landbot agents that talk to visitors one question at a time. For example:

● A lead qualification agent that asks a few questions and passes promising leads to your sales team.

● A form that asks one question at a time and saves the answers in a Google Sheet.

● A booking flow that offers a Calendly meeting slot.

● A support agent that answers questions from your help center with an AI Agent block and hands over to a person when needed.

What is the difference between the Landbot MCP and the Landbot Plugin?

The Landbot MCP server is the connection between an AI tool and your Landbot account. MCP stands for Model Context Protocol, the standard AI tools use to connect to other apps. The Landbot Plugin is a package for Claude Code and the Claude desktop app that includes that same connection and adds the Landbot Skills: Landbot's own instructions for building, styling and testing agents.

● The Landbot MCP is a connection you add to your AI assistant. In Claude it's called a connector. It works like connecting an app to your Google account: you add it once, sign in to Landbot, and the AI assistant can work in your Landbot account. Nothing is installed on your computer, and there is no API token to copy.

● The Landbot Plugin is an add-on you install in Claude Code or the Claude desktop app. It connects to the same Landbot MCP with the same sign-in, and adds a guided build: the agent is built, and styled if you describe a look, before it goes live; it's published after one approval and, in the Claude desktop app, tested on every path. Codex and Cursor can use the same two skills, copied from the plugin's repository.

Should you use the Landbot MCP or the Landbot Plugin?

Use the Landbot MCP to build and change agents from AI assistants that can connect to remote MCP servers with a sign-in, including Claude on the web and in the Claude mobile apps. Use the Landbot Plugin in Claude Code or the Claude desktop app when you want the guided build, a design written from your description and every path tested for you.

Landbot MCP

Landbot Plugin

What it is

A connection to your Landbot account (a connector, in Claude)

The Landbot MCP plus the Landbot Skills

Setup

Add the connection and sign in to Landbot

Install the plugin and sign in to Landbot

Build, change and publish agents

Yes

Yes. Styled before it goes live if you describe a look, then one approval to publish

Add an AI Agent block and teach it from texts and web pages

Yes

Yes

Connect Google Sheets, SendGrid and Calendly

Yes

Yes

Change how the web chat looks

Yes, with the CSS your AI assistant writes

Yes, written from your description with Landbot's styling guide and checked on the share link once it's live

Test every path of the agent for you

No

Yes, in the Claude desktop app

Messaging-app and step-form looks

No

Yes, as code you paste in the Landbot builder

Upload a file from your computer for the agent to show

Yes, from a tool that can run commands on your computer, such as Claude Code or Codex

Yes

Works in

Claude (web, desktop app, Cowork and the mobile apps), Claude Code, Codex, Cursor, VS Code with GitHub Copilot, Windsurf, Kiro, and app builders such as Lovable, Replit and Base44. Not ChatGPT yet.

Claude Code and the Claude desktop app (Code tab, Cowork). The two skills also run in Codex and Cursor.

What do the Landbot MCP and the Landbot Plugin change in your Landbot account?

The Landbot MCP and the Landbot Plugin ask for your approval before publishing and before changing an agent you already have.

● Before changing an agent you already have, the Landbot MCP and the Landbot Plugin name the agent and ask for your approval.

● Publishing an agent always needs your approval. Visitors keep the current version until you approve, and a yes covers one publish: a later change needs another yes.

● The Landbot MCP and the Landbot Plugin sign in through Landbot in your browser. You never type a password, API key or token into the chat.

● The sign-in screen lists what the AI tool may do: read, create, edit and publish your agents, read, create and edit their AI agents, read your connected accounts and the spreadsheets they reach, ask for another account to be connected, and read and change how your web chats look. It only works in the Landbot account you signed in to, with the permissions your Landbot user already has.

Where does your data go?

The Landbot MCP and the Landbot Plugin send requests only to Landbot's MCP server, mcp.landbot.io, which acts on your Landbot account with the access you approved at sign-in. To check a published agent, the Landbot Plugin also reads Landbot's public share pages.

For privacy details, read the Landbot Plugin's Privacy policy and Landbot's Privacy Policy.

Where can developers learn more?

The Landbot MCP and the Landbot Plugin run on Landbot's public APIs. Developers can read the Landbot MCP server guide, which lists every tool and permission, the Bots API concepts and the Bots API reference to build their own integration.

Was this article helpful?

How to build, test and publish an agent with the Landbot MCP or the Landbot Plugin

Contact