Slaide ships an MCP server. Any MCP client can use it. Claude, Codex, Gemini, or your own agent. It can scaffold, validate, and render decks. The server runs the same @aivorynet/core engine as the CLI, so there is no drift.
slaide mcp # start the server, stdio transport
Point your MCP client at that command. A typical config:
{
"mcpServers": {
"slaide": { "command": "npx", "args": ["@aivorynet/slaide", "mcp"] }
}
}
Tools
| Tool | Purpose |
|---|---|
slaide_scaffold(title, subtitle?, outline?, master?) |
Generate starter deck text |
slaide_validate(source | path) |
Validate and return diagnostics |
slaide_render_html(source | path, outPath) |
Render a deck to an HTML file |
slaide_render_pdf(source | path, outPath) |
Render a deck to a PDF file |
slaide_list_themes() |
List available masters and their layouts |
Resources
| Resource | What it serves |
|---|---|
slaide://spec |
The full language spec, in Markdown |
slaide://themes |
The theme guide and layout catalog |
Prompts
| Prompt | Purpose |
|---|---|
new_deck(topic) |
Reads the spec, scaffolds a deck, validates it, and renders it |
The spec and theme guide are exposed as resources. So an agent reads the language on demand instead of guessing. That is a big reason models write Slaide reliably.
Want an in-editor setup instead? Use the agent skill.