MCP Server
Connect ChatGPT, Claude, and other AI assistants to Encelade through the Model Context Protocol — list, generate, and manage presentations from natural-language prompts.
Encelade exposes a Model Context Protocol (MCP) server so AI assistants can interact with your workspace directly. Once connected, ChatGPT, Claude, or another MCP-compatible client can browse your projects, generate full presentations, and update or delete decks — all from natural-language prompts.
What you can do
- Browse decks — list projects and inspect their pages, theme, and collaborators.
- Generate a presentation — produce a deck, one-pager or proposal from a brief; poll the run for progress.
- Update or delete — rename projects, change themes, or permanently remove decks.
Supported clients
The MCP server speaks standard streamable HTTP and OAuth 2.0, so any MCP-compatible client can connect:
- ChatGPT (approved Encelade app in the directory)
- Claude on the web and desktop (custom connector)
- Claude Code (
claude mcp add) - Codex (
codex mcp add) - Any other client that supports streamable HTTP MCP with OAuth 2.0 discovery
For clients that don't yet speak streamable HTTP, the same endpoint also serves an SSE fallback at /api/mcp/sse (with messages posted to /api/mcp/message). Prefer streamable HTTP when both are supported.
Endpoint
https://www.encelade.ai/api/mcpAuthentication
Every request requires a Bearer token. Clients that support OAuth 2.0 discovery (RFC 9728) — including ChatGPT, Claude, Claude Code, and Codex — pick up the authorization server automatically and walk you through a browser-based authorize step. You sign in, pick a workspace, and approve the connection — there is no token to paste. Other clients can send an API token as a header instead.
Follow the setup tutorial to connect Encelade to ChatGPT or Claude, authorize your workspace, and test the connection.