Runway brings generative video to Claude Code and Cursor

Adding generative video to your product used to be a chore. You had to log into a dashboard, manage API keys, copy-paste tokens into your env file, and write boilerplate while juggling documentation tabs.

Runway is trying to hand that entire headache over to coding agents.

Runway launched Runway Dev plugins for Claude Code, Codex, and Cursor. With a single install, your coding agent hooks into your Runway Dev account via MCP and picks up the skills needed to build with official SDKs without ever leaving your editor.

Inside your agent's new media toolkit

The setup handles the boring admin stuff right away. Once connected, your agent automatically inspects your Runway Dev account, checking which projects are active, which models are available, pricing tiers, rate limits, and remaining credits. It sizes the integration strictly to what your project can access.

When you describe what you want to build, the agent takes over the heavy lifting. It can spin up a Model Router inside your project to dynamically send tasks to the right model, or configure a Character component for real-time avatars.

For standard pipelines, like generating a short product video whenever a new item hits an e-commerce catalog, the agent can build on top of Recipes, which are Runway's pre-made pipelines. It can also hook straight into Workflows that your team previously published as API endpoints in the Runway app.

Debugging gets a lot smarter too. If an asset generation pipeline fails, the agent checks the underlying error logs and inspects why the Model Router made a specific routing decision, figuring out what to adjust next on its own.

Best of all, security is clean: you sign in with your Runway Dev account on first use, and no API keys are stored inside the plugin.

Getting started takes just a quick prompt or terminal line, whether you are running Claude Code plugins, Cursor Agent chat, or Codex.

Why it matters

We think Runway is making a very smart bet here. The developer experience is shifting rapidly from writing client libraries by hand to orchestrating autonomous agents, and Runway is meeting builders where they actually spend their time.

For indie hackers and product engineers, this lowers the bar to shipping rich media features dramatically. Instead of spending hours reading API specs and wiring up routing logic, you can simply tell your agent what the user-facing feature should do. The agent checks your balance, sets up the endpoint, and wires the pipeline end-to-end.

For the broader AI ecosystem, it shows how MCP is quickly becoming the default bridge between agents and external infrastructure. The real battle among frontier model providers is no longer just raw benchmark performance, but how effortlessly an agent can discover, configure, and operate their platform.