Incredible.dev

Developer content made from the material you already have. Bring a repository, docs, a blog post, a talk or your notes, and a team of AI agents researches it, storyboards it and makes the film, the thumbnails, the cards and the posts, in channels you can steer and on a canvas you can review.

A new beginning. This is the new Incredible.dev. The earlier video creator lives on the archive/dev-video-creator branch.

How it flows

1. Start from your material

Ask in a channel and attach what you have, or place a template: Launch kit, Explainer, Post to social, Ad set, Pitch pack, Social pack or Talk to clips.

2. Research your code

A research frame asks local coding agents (Claude Code, Codex, Kimi Code) about your repository. Keep any passage of an answer as a note: each note records where it came from, and notes feed what gets made.

3. Storyboard, then sign off

Films start as a storyboard. Pin a note to a frame and the next version answers it; approve when it's right.

4. The film and everything around it

The film is built in code with HyperFrames, then thumbnails, social cards and GIFs, each linked on the canvas to what it was made from.

5. Steer it in a thread

Every task is a thread with its own lead agent and computer. Reply to steer the work, approve a preview, or stop it.

Your brand on everything

Read a brand from a website, or keep a brand pack, and choose which one each channel uses.

Run it locally

You need Docker, uv and Node.js.

uv sync --all-extras --dev
uv run design-agent sandbox build                    # the agents' computer
(cd web && npm install && npm run build)             # the app
uv run python -m design_agent.surfaces.web --demo    # a scripted lead, no API keys
uv run python -m design_agent.surfaces.web           # a real model (a provider key in .env)

Then open http://127.0.0.1:8765. For app development, run npm run dev in web/ (port 5178); it passes /api through to the server. Models lists the providers and their keys.

The app has no sign-in yet. It serves 127.0.0.1 only and takes the person from a request header, so don't expose it on a network.

How it's built

• The agents. Each task gets a lead agent that plans and hands work to helper agents, a Docker sandbox of its own, skills for each kind of work, and checks it runs on its renders (brand, motion, output). A cost meter tracks every run.
• The app. The server in src/design_agent/surfaces/web/ (API, live events, store, canvas tools, research, brands) and the web app in web/ (React, Vite, TypeScript), plugged in through one presenter hook, src/design_agent/surfaces/presenter.py.

More: architecture · instructions.md · skills/ · customization · changelog.

License

Apache 2.0. Third-party notices are in NOTICE.md.