Rive News · June 12, 2026
Rive Now Connects to AI Tools via MCP
Rive Editor now connects to Claude Code, Cursor and Codex through MCP. What that gives you over the built-in agent, the limits, and how to set it up in two minutes.
Rive added MCP (Model Context Protocol) support to the desktop editor. In plain terms: the AI agent you already pay for (Claude Code, Cursor, Codex) can now open your Rive file and work in it, with the same abilities as the agent built into the editor. Read the file, edit scripts, create View Models, wire State Machines, export. The editor stays the editor. What changes is who is driving.
What Changed?
Any agent that speaks MCP can connect to the Rive editor. Claude Code and Cursor are the two Rive documents; Codex is announced as supported.
It handles the repetitive wiring in one pass: hundreds of states, inputs and listeners, a View Model with all its properties, a layout.
Available in the Mac desktop app. Windows is planned.
Why connect your own agent?
I built most of the Scripting Course with Rive's built-in agent: 18 scripts, about 4,600 lines of Luau. It's good, and every prompt in the course still runs on it. I switched the rest of the course to my own agent through MCP for four reasons.
One bill, not two. Rive's agent runs on credits you top up inside the editor. If you already pay for Claude, Cursor or Codex, that subscription drives Rive too.
Faster on a real project. Once a script passes a few hundred lines, every change takes the built-in agent longer. A connected agent edits the ten lines that need changing and is done.
It remembers. Rive's agent starts every session blank. Yours keeps a project folder, a notes file, the values you tuned last week.
It's not limited to Rive. Your agent can export the file, upload it to your site, update the runtime, fix the code that loads it. Rive's agent stops at the editor's edge.
The honest limits
It's a young integration. A new script can take a few seconds to appear. Editor screenshots don't run scripts, so you verify on screen. Rive has to be open with your file open, and you start the agent after Rive so it finds the tools. Once, mid-session, Rive crashed. None of it changed my mind; all of it is worth knowing before your first session.
Setting it up
Nothing to install on the Rive side. Open Rive with your file, then ask your agent to connect to the Rive editor's MCP server. Claude Code, Cursor or Codex takes it from there. The video above is Rive's own walkthrough with Claude Code.
What This Means for You
The agent speeds up the building. It does not replace understanding. You still need to know how state machines, inputs and data binding work to direct it well and to catch its mistakes. The prompt skill is the same in both worlds: describe what you want to see, say what's wrong on screen, protect what already works. That is exactly what the Rive Scripting Course teaches, and from the Explosion module on, the lessons are recorded with an agent connected through MCP, with a skill file included so your agent starts every session already knowing how we work in Rive.
The full story, with the moment I switched: Rive's AI Agent or Your Own? Connecting Claude, Cursor and Codex to Rive.