Codex can now wire Runway’s video API into your app from one plugin

Runway said in a post on X that Runway Dev is now available on Codex. According to the company, one plugin install gives the coding agent what it needs to build Runway’s API into an app, with Codex doing the integration on its own.

The post, as visible in the announcement text, reads: “Runway Dev is now on Codex One plugin install gives Codex everything it needs to build Runway’s API into your app. Describe what your app should generate and let Codex build the integration autonomously.” The text we had access to is cut off at the link where developers can get started.

One install replaces hand-wired API setup

The idea is a change in who does the plumbing. A developer describes what the app should generate, for example video or images, and Codex handles the work of calling Runway’s models. Runway’s post does not list what the plugin contains, which Codex versions or plans support it, or whether it costs anything beyond normal Runway Dev usage.

This follows Runway’s earlier move for coding agents. On September 2, 2026, the company introduced Runway Dev MCP, a hosted MCP server that links a Runway Dev account to Claude, ChatGPT, Codex and Cursor. Users add it as a custom connector at https://dev.runwayml.com/mcp and approve access. Runway said at the time that an agent can compare models by price, capabilities and accepted inputs, pull the request schema for a chosen model, set up a Model Router, manage Characters and diagnose failed generations.

It is not confirmed in the post how the new Codex plugin relates to that MCP server. It may package the same capabilities for one-step installation, but Runway has not said so in the text available.

What the source leaves unanswered

Several practical details are missing. The post gives no pricing, no regional availability and no sample of what Codex produces. The claim that Codex can build the integration autonomously is Runway’s own and has not been independently checked. Quality will depend on how well the agent handles real-world edge cases, such as rejected generations or oversized assets, which the MCP server was designed to surface.

Developers who want to try it need a Runway Dev account and API key, which Runway’s developer documentation points to at dev.runwayml.com. Those who prefer the earlier route can still use the MCP connector described above.

Guardrails matter when an agent writes the integration

Letting a coding agent choose models and write calls to a generative media API raises a question about control, and Runway’s earlier MCP announcement is the best guide to how the company answers it. In that post, Runway said Runway Dev includes spend management and per-model cost visibility, built-in content moderation and IP indemnification. It also cited 99.9% uptime, SOC 2 Type II compliance, and a vendor management program that every model provider passes.

Those are the company’s own statements about the platform, and the Codex post does not repeat them. They are relevant because an autonomous integration can route requests across several models, and Runway says its Model Router can optimize for cost, latency or quality. In the same MCP post, Runway said errors such as a moderation rejection can be traced to a specific cause, which gives a developer or the agent a way to debug instead of guessing.

The approach is also part of a broader push by Runway to put its models inside other companies’ assistants. On October 19 the company said Runway MCP was available in Grok Bot, and in the past week it has also promoted App Studio, a no-code tool for building creative apps, and Runway Ads. The Codex plugin addresses a different audience: engineers who want generation features inside their own products rather than inside Runway’s apps.

For readers following Runway’s video work, our earlier coverage of the Aleph model shows the kind of capability that developers could now reach through an agent-built integration, though the post does not say which models the plugin exposes.