PPT Agent is a multi-agent workflow designed for automated presentation generation, addressing the need for efficient and consistent slide creation. It targets users requiring rapid development of visually structured presentations from textual prompts. The primary benefit lies in its LLM-driven content generation combined with AI-powered quality review and a standardized SVG output, enabling cross-platform compatibility and predictable results.

How It Works

This project employs a sophisticated multi-agent architecture, breaking down the presentation generation process into distinct roles: research-core for information gathering, content-core for outlining and drafting, slide-core for SVG design using a Bento Grid layout, and review-core for quality assurance powered by Gemini. The workflow progresses through stages including requirement research, material collection, outline planning, draft generation, design, and final review, incorporating user approval checkpoints. This structured approach, coupled with a defined quality scoring system (weighted total, layout, readability), ensures a high standard of output.

Quick Start & Requirements

Installation is performed via the Claude plugin marketplace: claude plugin install ppt-agent

Usage involves a command-line interface within the Claude environment: /ppt-agent:ppt <主题或需求描述>

Customization options include --style (business, tech, creative, minimal), --brand-colors (path to a YAML file), and --pages (target page count). A brand.yaml file allows defining primary/secondary colors and logo text. The primary prerequisite is a compatible Claude environment.

Highlighted Details

• Multi-Agent Workflow: Dedicated agents for research, content, slide design, and review.
• Gemini-Powered Review: AI-driven quality control integrated into the generation pipeline.
• Bento Grid SVG Output: Generates standardized 1280x720 SVG slides with a distinct Bento Grid layout.
• Quality Scoring: Implements metrics for weighted score, layout, readability, and repair rounds.
• Cross-Platform Validation: Verified on platforms like OpenCode (GPT-5.4, MiniMax M2.5) and Claude Code (Claude Opus).
• Customizable Branding: Supports style presets and custom brand color/logo integration via YAML.

Maintenance & Community

The project outlines a long-term roadmap including MCP Serverization for broader tool integration across hosts (Claude Code, Codex, OpenCode, etc.), core decoupling for platform-independent logic, and Headless mode for CI/CD pipelines and API usage. No specific community links (Discord, Slack) or direct contributor information were provided in the README.

Licensing & Compatibility

The project is released under the MIT License. This license is highly permissive, generally allowing for commercial use, modification, and distribution without significant restrictions, making it suitable for integration into both open-source and closed-source projects.

Limitations & Caveats

The primary installation method via Claude plugin may limit accessibility for users not operating within that ecosystem. While cross-platform support is a stated goal and has seen validation on specific platforms, the roadmap indicates ongoing development towards broader compatibility, suggesting some features or integrations might still be maturing.