README.md•3.67 kB
# Createve.AI Nexus MCP Client Server
A TypeScript implementation of an MCP server that connects to a remote Createve.AI Nexus server and exposes its functionality through the Model Context Protocol, for local deployment in tools that cannot connect to a remote MCP server.
## Overview
This MCP client server acts as a bridge between AI assistants that support the Model Context Protocol and a remote Createve.AI Nexus server. It dynamically generates MCP tools and resources based on the API endpoints available on the Nexus server.
## Features
- **Local Execution**: Runs on the client machine, avoiding direct WebSocket connections to remote servers
- **API Abstraction**: Interacts with the remote Nexus server via REST APIs
- **Enhanced Security**: Keeps API keys local to the client machine
- **Dynamic API Discovery**: Automatically discovers available APIs through OpenAPI schema
- **LLM-Optimized Interface**: Formats responses in a way that's optimal for LLM consumption
## Installation
```bash
npm install @createveai/nexus-mcp-server
```
## Configuration
The MCP client server can be configured using environment variables or command-line arguments:
### Environment Variables
- `CREATEVEAI_NEXUS_BASE_URL`: The base URL of the Nexus server (required)
- `CREATEVEAI_NEXUS_API_KEY`: API key for authentication (required)
- `CREATEVEAI_NEXUS_API`: Optional API filter to limit exposed endpoints
- `DEBUG`: Set to "createveai-nexus-mcp" to enable debug logging
### Command-Line Arguments
- `--base-url <URL>`: The base URL of the Nexus server
- `--api-key <KEY>`: API key for authentication
- `--api <FILTER>`: Optional API filter
- `--debug`: Enable debug logging
## Integration with AI Assistants
### Claude Integration
Add the MCP server to your Claude environment by modifying the MCP settings configuration file:
```json
{
"mcpServers": {
"createveai-nexus": {
"command": "node",
"args": ["/path/to/node_modules/@createveai/nexus-mcp-server/build/index.js"],
"env": {
"CREATEVEAI_NEXUS_BASE_URL": "https://nexus.createve.ai",
"CREATEVEAI_NEXUS_API_KEY": "your-api-key-here"
},
"disabled": false,
"autoApprove": []
}
}
}
```
## Available Resources
The MCP client server exposes the following resources:
- `queue://{queueId}`: Status and result for a queued request
- `schema://openapi`: OpenAPI schema for all available APIs
- `schema://{api}`: OpenAPI schema for a specific API (when using API filter)
- `docs://readme`: Server documentation
- `docs://{path}`: Specific documentation resource
## Available Tools
The MCP client server dynamically creates tools based on the API endpoints available on the Nexus server. These tools follow the naming convention of the API paths with slashes replaced by underscores.
For example, an API endpoint at `/api/text_processing/analyze` would be available as the MCP tool `text_processing_analyze`.
## Development Information
### Build Process
This package includes pre-built JavaScript files in the repository to ensure reliable packaging and distribution. The TypeScript source is compiled to JavaScript during development, and the built files are committed to the repository.
If you make changes to the TypeScript source:
1. Run `npm run build` to compile the TypeScript code
2. Commit both the source changes and the updated build files
### GitHub Workflow
The GitHub Actions workflow for this package relies on the pre-built files being present in the repository. It verifies their existence during the publishing process but doesn't attempt to rebuild them.
## License
This project is licensed under the MIT License.