Agents

Have your agent use Decodx

decodx is built for agents. Every command is non-interactive and every artifact is schema-validated JSON, so an agent can author, render, and revise a video end to end — over MCP or the CLI.

Read the MCP guide

Skills — create and edit video

Give your agent a video skill. It authors the spec, renders it, and reads back a structured result — no screen, no clicking.

Create from a brief

From a plain-English brief, the agent drafts a scene spec and renders a finished video.

Edit by editing the spec

A headline, a clip, a duration — the agent changes text, not a timeline.

Re-render for free

Content-hash caching means the agent can iterate dozens of times for seconds each.

Explore video skills

MCP — connect your assistant

decodx's MCP server exposes the video engine to any MCP-capable agent. Describe the video in plain language; the agent does the rest.

Works with your assistant

Connect Claude, ChatGPT, or Gemini and make videos from the chat you already use.

Schema-validated tools

Every tool takes and returns validated JSON, so the agent gets errors that teach the fix.

Fully editable output

The result is a real decodx spec — anything the agent made, a human can refine.

See the MCP page

Coming soon

Autonomous agents

Today your agent drives decodx over MCP. Next, it will own the whole job — planning, rendering, reviewing, and revising a video from a single goal.

See what's coming

An agent proposes; it never surprises

Paid work is estimated up front and gated behind approval, so autonomous work never runs up a bill.

Start free trial