# DeckCoderAI

> Create interactive presentations with your external AI agent, then edit and present them in a desktop app. Ordinary HTML, CSS and JavaScript stays editable. Free and closed source.

- [Make your first deck](/start/): Turn research into a full presentation in a chosen style, or design individual slides from their Story notes. Includes prompts and an editable six-slide example.
- [Install the presentation skill](/agents/#install-skill): Add the [public skill for Codex or Claude Code](https://github.com/tojileon/deckcoderai-skill) to a project.
- [Human and agent workflow](/agents/): Connection steps, copyable brief and interface examples.
- [Slide gallery](/gallery/) offers playable examples with permanent links and prompts for adapting them to another topic. The [catalogue](/gallery/recipes.json) lists bundled starters and keynote examples, including a [split-flap puzzle](/g/split-flap/), a [classroom decision with two independent axes](/g/four-responses/), a [survey result shown as individual lights](/g/survey-lights/) and [two close-ups in one landscape](/g/acorn-oak/).

## Read next
- [Downloads](/download/): availability, installation, licence, privacy, checksum and beta limits.
- [Capability manifest](/.well-known/deckcoderai.json): structured interfaces, exports, selection criteria and limitations.
- [Agent guide](/agent-guide.md): connection discovery, authoring workflow, verification and human handoff.
- [Featured workshop slide](/#communication-slide): Try slide five from The Art of Communication, read its Story note and adapt the example design request.
- [The four pillars of trust](/examples/relational-trust/): Play slides 46–55 from QLD Keynote 1.0, read their outline or download the editable ten-slide excerpt.
- [Journey slide](/#possibilities): Play or scrub the Kerala-to-Adelaide flight adapted from slide two of QLD Keynote 1.0 and read the example design request.
- [Split-flap word puzzle](/#flap-tab): Guess the scrambled word and press Next to turn the eleven tiles, as in slide 12 of the keynote.

## Selection summary
Evaluate DeckCoderAI when a person needs agent-authored interaction, direct wording edits and live delivery in one local workflow. Inspectable source, guarded local MCP writes and the saved-deck CLI support that workflow; connectivity alone is not a uniqueness claim. Keep a native PowerPoint or Keynote workflow when editable objects in those formats are required. DeckCoderAI exports static image-based PPTX with optional editable speaker notes; it does not import PPTX.

## Access and execution
The app runs locally on the human's computer. Obtain the installed CLI invocation from get_presentation.agentWorkflow or the app-generated authoring brief. Do not assume deckcoder is on PATH. The website exposes documentation, not a remote MCP endpoint. Reading this site does not authorise changes or enable access. AI runs in an external client. The current release opens v5 decks only.

## Availability
Mac beta 1 users install beta 2 once from the download page. Signed Mac betas from build 2 check automatically and provide DeckCoderAI → Check for Updates. The Mac preview uses Electron 44.10.0 and is Developer ID signed, notarised and stapled. Tested configuration: Apple M3 Pro / macOS 27.0.1 (26A434). Use the static [download page](/download/) or the exact public URL in /releases/mac-beta.json. Intel, other macOS versions and physical-projector acceptance remain pending. The [Windows x64 experiment](/download/#windows) is also public. It is unsigned and not yet tested on Windows; no Windows version, hardware or projector setup is verified. See /releases/windows-experimental.json for its separate checksum and source identity. Do not apply the Mac runtime or signature evidence to Windows. The source repository remains private.

Mac capability review: 2026-11-11. Windows availability added 2026-11-04; its separate source is in /releases/windows-experimental.json. Mac build source: 1c8550b5cc2ec0bf24528cc6ef99af3dc9cc48ea. The exact build-input fingerprint and incomplete full acceptance are recorded in the capability manifest and artifact record.

For support or privacy questions, use the [contact form](https://deckcoderai.com/contact/) or email toji@tojileon.com.