# Build MCP Servers for AI Tools

Building reliable MCP servers requires protocol knowledge, secure API design, and strong evaluation practices. This skill provides structured guidance for planning, coding, testing, and improving MCP integrations.

## Install

```bash
npx skillstore add anthropics/mcp-builder
```

## Metadata

- Status: approved
- Slug: mcp-builder
- Skillstore revision: r1
- Version status: missing
- Tree hash: 88d0330cb843321d7dee80a53cff04d01bc9328cb96cc603f2ae28bb6a0713c6
- Author: anthropics
- GitHub username: anthropics
- License: Complete terms in LICENSE.txt
- Repository: https://github.com/anthropics/skills/tree/main/skills/mcp-builder
- Ref: b8ca75d2c0a7e7102978993058777d82b8ab2610
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, env\_access, external\_commands
- Quality score: 75
- Public page: https://skillstore.pages.dev/skills/mcp-builder
- Manifest: https://skillstore.pages.dev/api/skills/mcp-builder/manifest

## Capabilities

- Explains MCP server planning, naming, transport selection, and tool design.
- Provides TypeScript and Python implementation patterns for MCP tools.
- Shows how to validate inputs with Zod or Pydantic schemas.
- Includes guidance for pagination, structured output, and error handling.
- Supplies an evaluation workflow and scripts for testing MCP servers.
- Links to protocol and SDK documentation that developers should review.

## Use Cases

- Create a New MCP Integration: Plan and implement an MCP server that exposes a product API through well-described tools.
- Improve Existing Tool Design: Review tool names, schemas, response formats, and annotations for better agent usability.
- Evaluate Server Quality: Create evaluation questions and run the included harness against an MCP server.

## Prompt Templates

### Plan a Basic MCP Server

```
Help me plan an MCP server for this API: [describe service]. Identify core tools, naming patterns, transport choice, and required authentication.
```

### Design Tool Schemas

```
Design MCP tool schemas for these operations: [list operations]. Include validation rules, response formats, pagination, and clear descriptions.
```

### Review an MCP Implementation

```
Review my MCP server implementation for protocol fit, input validation, error handling, response size, and security issues. Focus on actionable fixes.
```

### Create Evaluation Tasks

```
Create evaluation questions for this MCP server. Make them read-only, realistic, stable, independently verifiable, and difficult enough to test tool quality.
```

## Limitations

- It is a development guide, not a complete server generator.
- External API details must be supplied by the developer.
- Evaluation scripts require a configured Anthropic API key.
- Generated examples still need security review before deployment.

## Best Practices

- Start with protocol and SDK documentation before writing server code.
- Use strict input schemas and clear tool descriptions for every tool.
- Evaluate with realistic read-only tasks before publishing the server.

## Anti Patterns

- Do not expose every API endpoint without considering agent workflows.
- Do not return unbounded data without pagination or truncation.
- Do not place secrets in code, examples, logs, or tool responses.

## Security Audit

- Audited at: 2026-07-06T18:06:25.841\+00:00
- Summary: All static findings were reviewed against the referenced files and are false positives. The matches are documentation examples, placeholder credentials, template literals, or user-invoked evaluation harness options; no prompt injection attempt or malicious behavior was found.

## Stats

- Views: 244
- Downloads: 15
- Favorites: 0
- Popularity score: 0