docs/mcp/mcp-server-development-protocol.mdx
Build and share your MCP servers with the world. Once you've created a great MCP server, submit it to the Cline MCP Marketplace to make it discoverable and one-click installable by thousands of developers.
Model Context Protocol (MCP) servers extend AI assistants like Cline by giving them the ability to:
Without MCP, AI assistants are powerful but isolated. With MCP, they gain the ability to interact with virtually any digital system.
The heart of effective MCP server development is following a structured protocol. This protocol is implemented through a .clinerules file that lives at the root of your MCP working directory (/Users/your-name/Documents/Cline/MCP).
.clinerules FilesA .clinerules file is a special configuration that Cline reads automatically when working in the directory where it's placed. These files:
Here's the complete MCP Server Development Protocol that should be placed in your .clinerules file:
# MCP Server Development Protocol
CRITICAL: DO NOT USE attempt_completion BEFORE TESTING
## Step 1: Planning (PLAN MODE)
- What problem does this tool solve?
- What API/service will it use?
- What are the authentication requirements?
□ Standard API key
□ OAuth (requires separate setup script)
□ Other credentials
## Step 2: Implementation (ACT MODE)
1. Bootstrap
- For web services, JavaScript integration, or Node.js environments:
```bash
npx @modelcontextprotocol/create-server my-server
cd my-server
npm install
```
- For data science, ML workflows, or Python environments:
```bash
pip install mcp
# Or with uv (recommended)
uv add "mcp[cli]"
```
2. Core Implementation
- Use MCP SDK
- Implement comprehensive logging
- TypeScript (for web/JS projects):
```typescript
console.error("[Setup] Initializing server...")
console.error("[API] Request to endpoint:", endpoint)
console.error("[Error] Failed with:", error)
```
- Python (for data science/ML projects):
```python
import logging
logging.error('[Setup] Initializing server...')
logging.error(f'[API] Request to endpoint: {endpoint}')
logging.error(f'[Error] Failed with: {str(error)}')
```
- Add type definitions
- Handle errors with context
- Implement rate limiting if needed
3. Configuration
- Get credentials from user if needed
- Add to MCP settings:
- For TypeScript projects:
```json
{
"mcpServers": {
"my-server": {
"command": "node",
"args": ["path/to/build/index.js"],
"env": {
"API_KEY": "key"
},
"disabled": false,
"autoApprove": []
}
}
}
```
- For Python projects:
```bash
# Directly with command line
mcp install server.py -v API_KEY=key
# Or in settings.json
{
"mcpServers": {
"my-server": {
"command": "python",
"args": ["server.py"],
"env": {
"API_KEY": "key"
},
"disabled": false,
"autoApprove": []
}
}
}
```
## Step 3: Testing (BLOCKER ⛔️)
<thinking>
BEFORE using attempt_completion, I MUST verify:
□ Have I tested EVERY tool?
□ Have I confirmed success from the user for each test?
□ Have I documented the test results?
If ANY answer is "no", I MUST NOT use attempt_completion.
</thinking>
1. Test Each Tool (REQUIRED)
□ Test each tool with valid inputs
□ Verify output format is correct
DO NOT PROCEED UNTIL ALL TOOLS TESTED
## Step 4: Completion
❗ STOP AND VERIFY:
□ Every tool has been tested with valid inputs
□ Output format is correct for each tool
Only after ALL tools have been tested can attempt_completion be used.
## Key Requirements
- ✓ Must use MCP SDK
- ✓ Must have comprehensive logging
- ✓ Must test each tool individually
- ✓ Must handle errors gracefully
- NEVER skip testing before completion
When this .clinerules file is present in your working directory, Cline will:
Creating an MCP server requires just a few simple steps to get started:
.clinerules file (IMPORTANT)First, add a .clinerules file to the root of your MCP working directory using the protocol above. This file configures Cline to use the MCP development protocol when working in this folder.
Begin your Cline chat by clearly describing what you want to build. Be specific about:
For example:
I want to build an MCP server for the AlphaAdvantage financial API.
It should allow me to get real-time stock data, perform technical
analysis, and retrieve company financial information.
Cline will automatically start in PLAN MODE, guiding you through the planning process:
When ready, switch to ACT MODE using the toggle at the bottom of the chat to begin implementation.
One of the most effective ways to help Cline build your MCP server is to share official API documentation right at the start:
Here's the API documentation for the service:
[Paste API documentation here]
Providing comprehensive API details (endpoints, authentication, data structures) significantly improves Cline's ability to implement an effective MCP server.
In this collaborative phase, you work with Cline to design your MCP server:
Once planning is complete, Cline helps implement the server:
Let's walk through the development process of our AlphaAdvantage MCP server, which provides stock data analysis and reporting capabilities.
During the planning phase, we:
We began by bootstrapping the project:
npx @modelcontextprotocol/create-server alphaadvantage-mcp
cd alphaadvantage-mcp
npm install axios node-cache
Next, we structured our project with:
src/
├── api/
│ └── alphaAdvantageClient.ts # API client with rate limiting & caching
├── formatters/
│ └── markdownFormatter.ts # Output formatters for clean markdown
└── index.ts # Main MCP server implementation
The API client implementation included:
Key implementation details:
/**
* Manage rate limiting based on free tier (5 calls per minute)
*/
private async enforceRateLimit() {
if (this.requestsThisMinute >= 5) {
console.error("[Rate Limit] Rate limit reached. Waiting for next minute...");
return new Promise<void>((resolve) => {
const remainingMs = 60 * 1000 - (Date.now() % (60 * 1000));
setTimeout(resolve, remainingMs + 100); // Add 100ms buffer
});
}
this.requestsThisMinute++;
return Promise.resolve();
}
We implemented formatters to display financial data beautifully:
/**
* Format company overview into markdown
*/
export function formatStockOverview(overviewData: any, quoteData: any): string {
// Extract data
const overview = overviewData
const quote = quoteData["Global Quote"]
// Calculate price change
const currentPrice = parseFloat(quote["05. price"] || "0")
const priceChange = parseFloat(quote["09. change"] || "0")
const changePercent = parseFloat(quote["10. change percent"]?.replace("%", "") || "0")
// Format markdown
let markdown = `# ${overview.Symbol} (${overview.Name}) - ${formatCurrency(currentPrice)} ${addTrendIndicator(priceChange)}${changePercent > 0 ? "+" : ""}${changePercent.toFixed(2)}%\n\n`
// Add more details...
return markdown
}
We defined five tools with clear interfaces:
server.setRequestHandler(ListToolsRequestSchema, async () => {
console.error("[Setup] Listing available tools")
return {
tools: [
{
name: "get_stock_overview",
description: "Get basic company info and current quote for a stock symbol",
inputSchema: {
type: "object",
properties: {
symbol: {
type: "string",
description: "Stock symbol (e.g., 'AAPL')",
},
market: {
type: "string",
description: "Optional market (e.g., 'US')",
default: "US",
},
},
required: ["symbol"],
},
},
// Additional tools defined here...
],
}
})
Each tool's handler included:
This critical phase involved systematically testing each tool:
{
"mcpServers": {
"alphaadvantage-mcp": {
"command": "node",
"args": ["/path/to/alphaadvantage-mcp/build/index.js"],
"env": {
"ALPHAVANTAGE_API_KEY": "YOUR_API_KEY"
},
"disabled": false,
"autoApprove": []
}
}
}
get_stock_overview: Retrieved AAPL stock overview information
# AAPL (Apple Inc) - $241.84 ↑+1.91%
**Sector:** TECHNOLOGY
**Industry:** ELECTRONIC COMPUTERS
**Market Cap:** 3.63T
**P/E Ratio:** 38.26
...
get_technical_analysis: Obtained price action and RSI data
# Technical Analysis: AAPL
## Daily Price Action
Current Price: $241.84 (↑$4.54, +1.91%)
### Recent Daily Prices
| Date | Open | High | Low | Close | Volume |
| ---------- | ------- | ------- | ------- | ------- | ------ |
| 2025-02-28 | $236.95 | $242.09 | $230.20 | $241.84 | 56.83M |
...
get_earnings_report: Retrieved MSFT earnings history and formatted report
# Earnings Report: MSFT (Microsoft Corporation)
**Sector:** TECHNOLOGY
**Industry:** SERVICES-PREPACKAGED SOFTWARE
**Current EPS:** $12.43
## Recent Quarterly Earnings
| Quarter | Date | EPS Estimate | EPS Actual | Surprise % |
| ---------- | ---------- | ------------ | ---------- | ---------- |
| 2024-12-31 | 2025-01-29 | $3.11 | $3.23 | ↑4.01% |
...
During development, we encountered several challenges:
Our AlphaAdvantage implementation taught us several key lessons:
Effective logging is essential for debugging MCP servers:
// Start-up logging
console.error("[Setup] Initializing AlphaAdvantage MCP server...")
// API request logging
console.error(`[API] Getting stock overview for ${symbol}`)
// Error handling with context
console.error(`[Error] Tool execution failed: ${error.message}`)
// Cache operations
console.error(`[Cache] Using cached data for: ${cacheKey}`)
Type definitions prevent errors and improve maintainability:
export interface AlphaAdvantageConfig {
apiKey: string
cacheTTL?: Partial<typeof DEFAULT_CACHE_TTL>
baseURL?: string
}
/**
* Validate that a stock symbol is provided and looks valid
*/
function validateSymbol(symbol: unknown): asserts symbol is string {
if (typeof symbol !== "string" || symbol.trim() === "") {
throw new McpError(ErrorCode.InvalidParams, "A valid stock symbol is required")
}
// Basic symbol validation (letters, numbers, dots)
const symbolRegex = /^[A-Za-z0-9.]+$/
if (!symbolRegex.test(symbol)) {
throw new McpError(ErrorCode.InvalidParams, `Invalid stock symbol: ${symbol}`)
}
}
Reduce API calls and improve performance:
// Default cache TTL in seconds
const DEFAULT_CACHE_TTL = {
STOCK_OVERVIEW: 60 * 60, // 1 hour
TECHNICAL_ANALYSIS: 60 * 30, // 30 minutes
FUNDAMENTAL_ANALYSIS: 60 * 60 * 24, // 24 hours
EARNINGS_REPORT: 60 * 60 * 24, // 24 hours
NEWS: 60 * 15, // 15 minutes
}
// Check cache first
const cachedData = this.cache.get<T>(cacheKey)
if (cachedData) {
console.error(`[Cache] Using cached data for: ${cacheKey}`)
return cachedData
}
// Cache successful responses
this.cache.set(cacheKey, response.data, cacheTTL)
Implement robust error handling that maintains a good user experience:
try {
switch (request.params.name) {
case "get_stock_overview": {
// Implementation...
}
// Other cases...
default:
throw new McpError(ErrorCode.MethodNotFound, `Unknown tool: ${request.params.name}`)
}
} catch (error) {
console.error(`[Error] Tool execution failed: ${error instanceof Error ? error.message : String(error)}`)
if (error instanceof McpError) {
throw error
}
return {
content: [
{
type: "text",
text: `Error: ${error instanceof Error ? error.message : String(error)}`,
},
],
isError: true,
}
}
Resources let your MCP servers expose data to Cline without executing code. They're perfect for providing context like files, API responses, or database records that Cline can reference during conversations.
server.setRequestHandler(ListResourcesRequestSchema, async () => {
return {
resources: [
{
uri: "file:///project/readme.md",
name: "Project README",
mimeType: "text/markdown",
},
],
}
})
server.setRequestHandler(ReadResourceRequestSchema, async (request) => {
if (request.params.uri === "file:///project/readme.md") {
const content = await fs.promises.readFile("/path/to/readme.md", "utf-8")
return {
contents: [
{
uri: request.params.uri,
mimeType: "text/markdown",
text: content,
},
],
}
}
throw new Error("Resource not found")
})
Resources make your MCP servers more context-aware, allowing Cline to access specific information without requiring you to copy/paste. For more information, refer to the official documentation.
Challenge: APIs often have different authentication methods.
Solution:
// Authenticate using API key from environment
const API_KEY = process.env.ALPHAVANTAGE_API_KEY
if (!API_KEY) {
console.error("[Error] Missing ALPHAVANTAGE_API_KEY environment variable")
process.exit(1)
}
// Initialize API client
const apiClient = new AlphaAdvantageClient({
apiKey: API_KEY,
})
Challenge: APIs may not provide all the functionality you need.
Solution:
Challenge: Most APIs have rate limits that can cause failures.
Solution:
if (this.requestsThisMinute >= 5) {
console.error("[Rate Limit] Rate limit reached. Waiting for next minute...")
return new Promise<void>((resolve) => {
const remainingMs = 60 * 1000 - (Date.now() % (60 * 1000))
setTimeout(resolve, remainingMs + 100) // Add 100ms buffer
})
}