Skip to main content

Overview

The Model Context Protocol (MCP) is an open standard for connecting Large Language Models (LLMs) to external platforms. ToolJet MCP is a server that exposes your ToolJet instance to an AI coding agent, so the agent can build and modify apps in the workspace you already work in, without you leaving your editor.

To connect your agent, follow the Setup guide.

What You Can Do​

Build Apps

Create an app from a prompt, with pages, datasource queries, and components generated and bound together for you.

Modify Apps

Change apps that already exist, whether built by hand, by ToolJet AI, or by your agent earlier. Components and queries are edited in place rather than recreated.

Work with Data

Inspect the datasources connected to your workspace, create ToolJet DB tables, seed them with rows, and write queries against ToolJet DB, PostgreSQL, REST APIs, or any other connected plugin.

Manage Permissions

Inspect, set, or clear server-enforced access on an individual page or query within an app.

Manage Users

Invite users to the workspace, update their profile and role, assign them to existing groups, and archive or unarchive them.

Interact in Natural Language

Describe the outcome you want in plain language. Your agent picks the right tools on its own, so you never name them yourself.

Benefits of Using ToolJet MCP​

  • No ToolJet credits: Operations run on your AI client's own model subscription, so building through MCP doesn't draw down your ToolJet AI credits.
  • Better output: Agents build against ToolJet's governed first-party contracts, so they work from the platform's real schemas instead of guessing at its internals.
  • Stays in your workflow: Create and modify apps from the editor you already work in, with no context switching into ToolJet.
  • Self-verifying: Agents can run their own queries to confirm the data is right, and those with browser access can open the app they built and check it works before handing it back.
  • Extensible: Pair ToolJet MCP with other skills, tools, and MCP servers to extend what your agent can do, such as reading a Figma design before building from it.

ToolJet MCP Auditability​

ToolJet MCP authenticates using a Personal Access Token (PAT). The PAT is created by a ToolJet user, scoped to a specific workspace, and operates within the permissions assigned to its owner. As a result, the MCP agent cannot access workspaces or resources beyond the token’s authorized scope.

Because MCP requests are authenticated through a user-linked PAT, changes made by the MCP agent are executed within the PAT owner’s ToolJet permission context. ToolJet’s audit-log system records platform activities, including the acting user, action performed, timestamp, location, and relevant application, query, user, group, or permission event.

Accordingly, MCP-driven changes that generate standard ToolJet platform events can be associated with the authenticated PAT owner in the audit trail. This provides accountability and traceability for changes made through ToolJet MCP.

Supported MCP Clients​

ToolJet MCP works with any MCP-compatible client. The following are supported directly, with a packaged plugin that installs the server and the tooljet-app-builder skill together:

Claude Code
Anthropic's coding agent in the terminal
Codex
OpenAI's coding agent in the CLI or IDE
Others
Any MCP-compatible client, such as Cursor or Cline

Clients outside this list connect by registering the bundled stdio server manually. See Setup for the configuration entry.

Requirements​

  • A ToolJet personal access token for the workspace you want to build in
  • Node.js 20 or newer, available to your AI client
  • An MCP-compatible AI client