decodx gives your AI agent a real video skill and an MCP server. Describe what you want — it authors the spec, renders the video, and writes the matching guide, the same way every time.
For your AI agent
Paste into Claude, Codex, Cursor, Gemini, or ChatGPT.
Drag to compare — raw capture vs. decodx render
For every team that has to explain software
Bring a screen recording, a slide deck, or clips you already have. Captures are cached, so you record once.
Scenes and narration live as plain text you can review in a pull request — no timeline to scrub.
decodx adds narration, captions, chapters, and branding, then renders the same way every time.
Change one scene; cached assets are reused and only what changed rebuilds — in seconds.
Author a video the way you author source. It lives in your repo and reviews like a diff.
The same spec and assets always produce the same MP4. No model in the render loop.
Content-hash caching means iterating costs seconds, not another export.
Accurate burned-in captions and chapter markers, generated from your narration.
The same render produces a written step-by-step article from the exact clips it used.
Non-interactive commands, schema-validated JSON, and spend gated behind approval.
Hand it to your agent or author it yourself — decodx renders it the same way, every time.