open-slide β€” The Slide Framework Built for AI Coding Agents

open-slide β€” The Slide Framework Built for AI Coding Agents

Most AI-generated slides are either ugly PowerPoints or limited Markdown renders. open-slide takes a different approach: it’s a React-based slide framework designed specifically for AI coding agents. Describe your deck in natural language, your agent writes React components on a standardized 1920x1080 canvas, and you get professional presentations with hot reload, in-browser commenting, presenter mode, and one-click export to HTML or PDF. Every slide is arbitrary code β€” versioned, reviewable, yours.

*Source: GitHub β€” 1weiho/open-slide open-slide.dev Documentation*

How It Works

You describe the deck in natural language
         ↓
Agent writes React components (one per slide)
         ↓
open-slide handles: canvas, scaling, navigation, hot reload
         ↓
You review in browser, click elements to add comments
         ↓
Agent applies comments via /apply-comments
         ↓
Export: static HTML site or print-ready PDF

The key insight: agents are good at writing React code, but bad at layout engines. open-slide gives agents a fixed canvas (1920x1080) and handles all the rendering infrastructure, so the agent focuses purely on content.

Agent Integration

Agent-agnostic framework that works with Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Windsurf, and Zed. The scaffold generates AGENTS.md + CLAUDE.md and versions skills in .agents/skills/. Ships with built-in skills:

Skill What It Does
/create-slide Generate a complete deck from a description
/slide-authoring Technical reference for writing slides
/apply-comments Apply all pending in-browser comments
/create-theme Generate a custom theme for brand consistency

Key Features

In-Browser Comment Inspector β€” Click any element in the dev server and attach a comment. Comments are persisted as @slide-comment markers in source code. Run /apply-comments and the agent applies every pending edit. This creates a tight human-in-the-loop feedback cycle without leaving the browser.

Assets Manager β€” Built-in panel for managing images, videos, and fonts per deck. Integrated svgl catalogue lets you search and drop in any brand logo with one click.

Presenter Mode β€” Fullscreen playback with keyboard navigation, plus a presenter view with current/next slide preview, speaker notes, and a timer. Built for the stage, not just the browser tab.

Slide Manager β€” Organize slides in folders with drag-and-drop reordering. Supports theme systems for brand consistency across decks.

Export β€” One command generates a self-contained static HTML site or a print-ready PDF. Deploy to Vercel, Cloudflare Pages, or Netlify.

Quick Start

# Scaffold a new project
npx @open-slide/cli init my-slide
cd my-slide

# Start dev server with hot reload
pnpm dev

# In Claude Code, generate slides
/create-slide "A 10-slide deck about transformer attention mechanisms"

# Export
pnpm build      # static HTML site
# PDF: use dev server toolbar β†’ Export β†’ PDF, or open with ?print=1 and print

Each slide is a React component in slides/<id>/index.tsx. Edit directly or let the agent handle it.

Architecture

Monorepo with pnpm + Turbo:

Package Purpose
packages/core Runtime, Vite plugin, dev/build/preview CLI
packages/cli Scaffolding (npx @open-slide/cli init)
apps/demo Reference workspace

Tech stack: TypeScript (95%), React, Vite, Vitest for testing.

open-slide vs Other Slide Tools

Tool Approach Agent-Ready Code Ownership
open-slide React components on fixed canvas Built-in skills Full (versioned .tsx)
reveal.js HTML/Markdown slides No native agent support HTML files
Slidev Markdown β†’ Vue slides Partial Markdown files
Google Slides / PowerPoint GUI-based API only Proprietary format
Marp Markdown β†’ PDF/HTML No Markdown files

open-slide’s advantage: slides are arbitrary React code, not constrained Markdown. You can embed interactive components, animations, data visualizations β€” anything React can render.

Real-World Use Cases

  • Conference talks β€” Describe your talk outline, agent generates professional slides with proper typography and layout. Refine interactively with click-to-comment.
  • Course materials β€” Faculty describe lecture content, agent produces slide decks with code examples, diagrams, and consistent branding.
  • Startup pitch decks β€” Iterate quickly on investor presentations with agent-generated content and human design feedback.
  • Technical documentation β€” Turn architecture docs into visual presentations for team onboarding or client demos.

How LearnAI Team Could Use This

  • Lecture slide production β€” Q can describe lecture topics and have Claude Code generate slide decks via /create-slide, then refine with the comment inspector.
  • Student presentation projects β€” Students learn React component patterns while building presentations β€” code literacy and communication skills together.
  • Agent tool design case study β€” open-slide’s click-to-comment β†’ /apply-comments loop is an excellent example of human-in-the-loop agent interaction design.