How does Deckpipe integrate with AI agents like Claude.ai or Cursor?
Deckpipe integrates with AI agents through a native remote server at /mcp or via a local installation using npx deckpipe-mcp. This allows agents to send JSON descriptions of decks to Deckpipe for rendering and management, and to interact with the Deckpipe API for various operations.
What kind of customization options are available for slide layouts and themes?
Deckpipe offers a variety of pre-defined layouts like title, bullets, SWOT, and charts. For theming, agents can specify any Google Font for headings and body text, choose an accent color for the entire deck, and add per-slide key takeaways. The icons_and_text layout also supports any Lucide icon or emoji.
Can I make real-time edits to a deck without regenerating the entire presentation?
Yes, Deckpipe supports live editing through PATCH requests with deep merge operations. This allows agents to perform granular updates such as inserting, deleting, moving, or replacing individual elements (like a single bullet point) without needing to resend the complete deck JSON.
How does Deckpipe handle image sourcing and attribution for presentations?
Deckpipe allows AI agents to search Unsplash for stock photos and drop them directly into slides. The platform automatically handles the necessary attribution and re-hosting of these images, simplifying the process for users and ensuring compliance.
What is the 'mcp' mentioned in the context of Deckpipe?
MCP stands for 'native remote server' and is Deckpipe's mechanism for AI agents like Claude.ai and Cursor to communicate with the platform. It provides the interface for agents to send commands and data to Deckpipe, enabling them to create, modify, and manage presentations programmatically.