Lucid

The Lucid MCP Server lets multimodal language models access and analyze Lucid diagrams by exporting visuals and metadata from Lucid apps. It wires Lucid document data into an MCP endpoint so you can ask natural-language questions, request AI-driven analysis, and retrieve lightweight document metadata for quick navigation.

How to use

You run the Lucid MCP Server locally and connect your MCP client to it over stdio. Start the server with your Lucid API key and optional AI provider keys. Then you can ask it to list your Lucid documents, fetch metadata, export diagram pages as images, or perform AI-enabled analysis on diagrams. Use natural-language prompts or call the server’s tools to discover documents, retrieve metadata, and analyze diagrams end-to-end.

How to install

Prerequisites: you need Node.js version 18 or higher and a Lucid API key.

1. Install the MCP server globally from npm.

npm install -g lucid-mcp-server

Configuration and run

Set the necessary environment variables in your shell before starting the server. At minimum you need the Lucid API key. You can also configure AI provider keys if you plan to run AI-powered analysis.

# Required for all features
export LUCID_API_KEY="your_api_key_here"

# Optional: For AI analysis, configure either Azure OpenAI or OpenAI

# Option 1: Azure OpenAI (takes precedence)
export AZURE_OPENAI_API_KEY="your_azure_openai_key"
export AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com"
export AZURE_OPENAI_DEPLOYMENT_NAME="gpt-4o"

# Option 2: OpenAI (used as a fallback if Azure is not configured)
export OPENAI_API_KEY="your_openai_api_key"
export OPENAI_MODEL="gpt-4o" # optional, defaults to gpt-4o

Note: Azure OpenAI keys take precedence if provided. If not, the server will fall back to OpenAI if an OpenAI key is configured.

## Starting the server

Run the server using the standard MCP runtime command. The server executable is named lucid-mcp-server and runs in the current shell with the configured environment.

lucid-mcp-server

## Verification

To verify the MCP server is reachable and ready for use, you can run a quick inspection command to test the integration. If you have the MCP Inspector tool available, run it against lucid-mcp-server to validate connectivity and capabilities.

## Additional configuration notes

The server supports both document discovery and lightweight tab metadata retrieval from Lucid Cloud accounts, plus PNG exports of diagrams and AI-powered analysis when AI keys are provided. Ensure your Lucid API key has permissions to access the documents you intend to query.

## Security considerations

Keep API keys secure and do not commit them to source control. Use environment-based configuration and restrict access to the MCP server to trusted clients. If you’re deploying this in a shared environment, consider using secret management to inject keys at runtime.

## Troubleshooting

If the server fails to start, verify that LUCID\_API\_KEY is set and that the command lucid-mcp-server is in your PATH. Check for any error messages related to OpenAI or Azure OpenAI configuration if AI features are not starting as expected.

## Contributors and tools

Key capabilities include document search, get-document metadata with optional AI analysis, and lightweight tab metadata access. You can also verify integration with MCP Inspector during development.

## Available tools

### search-documents

Lists documents in your Lucid account using search keywords to filter results.

### get-document

Fetches document metadata and can optionally perform AI analysis on the document's visual content.

### get-document-tabs

Retrieves lightweight metadata about all tabs/pages in a Lucid document for quick navigation.