# build-mcp

Comprehensive guide for creating high-quality MCP servers that enable LLMs to interact with external services through well-designed tools.

* Purpose - Build custom MCP servers for any service or API
* Output - Production-ready MCP server with tools and evaluations

```bash
/mcp:build-mcp
```

## When to Build Custom MCP Servers

Build an MCP server when you need the LLM to:

* Interact with internal company APIs or services
* Access databases or data sources not available via existing MCP servers
* Integrate with third-party services (CRMs, project management, communication tools)
* Perform specialized operations unique to your domain

## How It Works

The command guides you through a four-phase development process:

**Phase 1: Deep Research and Planning**

1. **Agent-Centric Design Principles**
   * Build workflow tools, not just API wrappers
   * Optimize for limited context windows
   * Design actionable error messages
   * Follow natural task subdivisions
2. **Protocol Study**: Load MCP specification from `modelcontextprotocol.io`
3. **Framework Selection**:
   * Python with FastMCP for rapid development
   * TypeScript with MCP SDK for type safety
4. **API Research**: Exhaustively study the target API documentation
5. **Implementation Planning**:
   * Tool selection and prioritization
   * Shared utilities design
   * Input/output schema design
   * Error handling strategy

**Phase 2: Implementation**

1. **Project Structure**: Set up according to language-specific best practices
2. **Core Infrastructure**: Build shared utilities first (API helpers, error handling, formatting)
3. **Tool Implementation**: Systematically implement each planned tool
4. **Annotations**: Add proper tool hints (readOnly, destructive, idempotent)

**Phase 3: Review and Refine**

1. **Code Quality Review**: DRY principle, composability, consistency
2. **Testing**: Verify syntax and imports (note: MCP servers are long-running, use evaluation harness)
3. **Quality Checklist**: Language-specific verification

**Phase 4: Create Evaluations**

1. **Tool Inspection**: Understand available capabilities
2. **Content Exploration**: Use read-only operations to explore data
3. **Question Generation**: Create 10 complex, realistic evaluation questions
4. **Answer Verification**: Verify each answer is correct and stable

## Usage Examples

```bash
# Start building an MCP server
> /mcp:build-mcp

# The command will guide you through:
# 1. Understanding your integration requirements
# 2. Choosing Python or TypeScript
# 3. Designing tools for your use case
# 4. Implementing with best practices
# 5. Testing and evaluation
```


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://cek.neolab.finance/plugins/mcp/build-mcp.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.