MCP Setup

Connect Cuppa Studio to your AI assistant — generate and refine presentations directly from Claude, Cursor, Windsurf, and more.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants use external tools. Cuppa Studio's MCP server gives any compatible AI client the ability to generate presentations, refine existing ones, list themes and templates, and more — all without leaving your AI workflow.

Prerequisites

# Set your API key (add to your shell profile for persistence)
export ANTHROPIC_API_KEY="sk-ant-..."

Setup by Client

Claude Desktop

Open Settings → Developer → Edit Config (or edit ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "cuppa-studio": {
      "command": "npx",
      "args": ["@cuppa-studio/mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-..."
      }
    }
  }
}

Restart Claude Desktop. You'll see Cuppa Studio tools in the tools menu (hammer icon).

Claude Code (CLI)

Add Cuppa Studio as an MCP server in your project or global settings:

# Add to current project
claude mcp add mycuppa -- npx @cuppa-studio/mcp

# Or add globally (available in all projects)
claude mcp add --scope user mycuppa -- npx @cuppa-studio/mcp

Alternatively, add it to your .mcp.json file in your project root:

{
  "mcpServers": {
    "cuppa-studio": {
      "command": "npx",
      "args": ["@cuppa-studio/mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-..."
      }
    }
  }
}

Cursor

Open Settings → MCP and click "Add new MCP server". Enter:

FieldValue
Namecuppa-studio
Typecommand
Commandnpx @cuppa-studio/mcp

Or add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "cuppa-studio": {
      "command": "npx",
      "args": ["@cuppa-studio/mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-..."
      }
    }
  }
}

Windsurf

Open Settings → Cascade → MCP and click "Add Server", or edit ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "cuppa-studio": {
      "command": "npx",
      "args": ["@cuppa-studio/mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-..."
      }
    }
  }
}

VS Code (Copilot)

VS Code supports MCP servers through GitHub Copilot's agent mode. Add to your .vscode/mcp.json:

{
  "servers": {
    "cuppa-studio": {
      "command": "npx",
      "args": ["@cuppa-studio/mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-..."
      }
    }
  }
}

Other MCP Clients

Any MCP-compatible client can use Cuppa Studio. The server runs over stdio — the universal configuration is:

SettingValue
Transportstdio
Commandnpx
Arguments@cuppa-studio/mcp
EnvironmentANTHROPIC_API_KEY=sk-ant-...

Verifying the Setup

After configuring, verify Cuppa Studio is connected by asking your AI assistant:

# Test the connection
"List all Cuppa Studio themes"

# Should return the 8 built-in themes with colors and fonts

If this works, you're ready to generate presentations. Try:

"Generate a tech talk about WebSockets with code examples and a mermaid diagram"

Available Tools

Once connected, your AI assistant has access to these tools:

ToolDescription
generate_presentationGenerate a presentation from a prompt — outputs a .md file
refine_presentationModify an existing presentation — reads .cup or .md, outputs .md
compile_markdownCompile Cuppa Studio markdown into a .cup file
read_cupfileRead and parse a .cup file
write_cupfileWrite a CupFile JSON object to a .cup archive
list_themesList all themes with color schemes and fonts
list_templatesList all presentation templates

Prompt Templates

The MCP server includes 10 pre-built prompt templates that structure your request automatically. Use them by asking your AI assistant to use a specific template:

# Using prompt templates
"Use the tech-talk template about WebSockets"
"Use the pitch-deck template for my startup Acme Corp"
"Use the incident-postmortem template for the March 5 outage"
TemplateInputThemeSlides
tech-talkTopicdark8-10
team-updateTeam namecorporate5-7
product-demoProductbold10-12
workshopTopicocean12-15
pitch-deckCompanyminimal10-12
architecture-reviewSystemdark8-10
sprint-demoSprintbold7-9
code-walkthroughTopicdark8-10
incident-postmortemIncidentdark7-9
api-docsAPI nameocean7-9

Workflow: MCP → Studio

The MCP server outputs .md files — standard Cuppa Studio markdown. Here's the recommended workflow:

  1. Generate — ask your AI assistant to create a presentation. It writes a .md file.
  2. Open in Studio — drag the .md file into Cuppa Studio or copy-paste the contents.
  3. Refine — edit the markdown directly in Studio, or ask the AI to refine it.
  4. Present — click Present in Studio, or use mycuppa serve slides.md.

Troubleshooting

ProblemSolution
Tools not showing up Restart your AI client after adding the config. Check that npx @cuppa-studio/mcp runs without errors in your terminal.
"ANTHROPIC_API_KEY not set" Add the env block to your MCP config, or export ANTHROPIC_API_KEY=... in your shell profile.
Slow first run The first npx invocation downloads the package. Subsequent runs are cached.
Permission errors Ensure Node.js 18+ is installed and npx is on your PATH.
Generated file not found By default, files are written to your system temp directory. Use the outputPath parameter to specify a location.

For prompt writing tips, see the AI Guide.