Nuclino MCP Server

A Model Context Protocol (MCP) server that provides access to Nuclino content through structured search and retrieval tools.

Features

• Search Nuclino content by team or workspace

• Retrieve specific items by ID

• Get team and workspace information

• Find teams and workspaces by name

• Rate limiting and retry logic for API reliability

• Support for both HTTP and stdio transports

Transport Types

This server supports two transport types:

HTTP Transport (Default)

• Best for web applications and remote access

• Supports multiple concurrent sessions

• Requires API key to be sent in headers for each request

• Runs on port 3000 by default

Stdio Transport

• Best for local development and CLI tools

• Single session per server instance

• Requires API key to be set as environment variable

• Communicates over stdin/stdout

Installation

npm install

Configuration

Environment Variables

• TRANSPORT_TYPE: Set to 'http' (default) or 'stdio'

• PORT: Port for HTTP transport (default: 3000)

• NUCLINO_API_KEY: Required for stdio transport only

HTTP Transport

1. Start the server:

npm run dev  # Development
npm run start  # Production

2. Connect to the server at http://localhost:3000/mcp

3. Include your Nuclino API key in the nuclino-api-key header with each request.

Stdio Transport

1. Set your API key:

export NUCLINO_API_KEY="your_api_key_here"

2. Start the server:

npm run dev:stdio  # Development
npm run start:stdio  # Production

The server will communicate via stdin/stdout, making it suitable for direct integration with MCP clients.

API Key

You need a Nuclino API key to use this server. You can create one in your Nuclino profile page.

Usage

You will find an .mcp.json.example file in this repository to show how you can configure both versions of the MCP. Once installed, you can use the MCP to summarize all info it can find on Nuclino about certain topics. You can stear it by passing a workspace name to limit the search scope.

Available Tools

• search_by_team: Search content within a specific team

• search_by_workspace: Search content within a specific workspace

• get_teams: Get all available teams

• get_workspaces: Get all available workspaces

• find_team_by_name: Find a team by name

• find_workspace_by_name: Find a workspace by name

• get_item: Get a specific item by ID

Development

Building

npm run build

Running in Development

# HTTP transport
npm run dev

# Stdio transport
npm run dev:stdio

Docker

# Build image
docker build -t nuclino-mcp-server .

# Run HTTP server
docker run -d -p 3000:3000 --name nuclino-mcp-server nuclino-mcp-server

# Run stdio server
docker run -e TRANSPORT_TYPE=stdio -e NUCLINO_API_KEY=your_key nuclino-mcp-server

Architecture

The server follows Clean Architecture principles:

• Domain Layer: Entities and repository interfaces

• Application Layer: Use cases and business logic

• Infrastructure Layer: External integrations (Nuclino API, transports)

• Presentation Layer: MCP server implementation

Transport Layer

The transport layer is abstracted to support multiple transport types:

• ITransport: Common interface for all transports

• HttpTransport: HTTP-based transport with session management

• StdioTransport: Stdio-based transport for local usage

• TransportFactory: Factory for creating transport instances

Error Handling

The server includes comprehensive error handling:

• Rate limiting with exponential backoff

• Retry logic for transient failures

• Graceful shutdown handling

• Detailed logging for debugging