Cline rules

1---
2description: Defines the protocol and steps for developing MCP (Model Context Protocol) servers.
3author: https://github.com/nickbaumann98
4version: 1.0
5category: "MCP Development"
6tags: ["mcp", "development", "protocol", "server", "integration"]
7globs: ["*"]
8---
9# MCP Server Development Protocol
10 
11⚠️ CRITICAL: DO NOT USE attempt_completion BEFORE TESTING ⚠️
12 
13## Step 1: Planning (PLAN MODE)
14- What problem does this tool solve?
15- What API/service will it use?
16- What are the authentication requirements?
17  □ Standard API key
18  □ OAuth (requires separate setup script)
19  □ Other credentials
20 
21## Step 2: Implementation (ACT MODE)
221. Bootstrap
23   - For web services, JavaScript integration, or Node.js environments:
24```bash
25     npx @modelcontextprotocol/create-server my-server
26     cd my-server
27     npm install
28```
29   - For data science, ML workflows, or Python environments:
30```bash
31     pip install mcp
32     # Or with uv (recommended)
33     uv add "mcp[cli]"
34```
35 
362. Core Implementation
37   - Use MCP SDK
38   - Implement comprehensive logging
39     - TypeScript (for web/JS projects):
40```typescript
41       console.error('[Setup] Initializing server...');
42       console.error('[API] Request to endpoint:', endpoint);
43       console.error('[Error] Failed with:', error);
44```
45     - Python (for data science/ML projects):
46```python
47       import logging
48       logging.error('[Setup] Initializing server...')
49       logging.error(f'[API] Request to endpoint: {endpoint}')
50       logging.error(f'[Error] Failed with: {str(error)}')
51```
52   - Add type definitions
53   - Handle errors with context
54   - Implement rate limiting if needed
55 
563. Configuration
57   - Get credentials from user if needed
58   - Add to MCP settings:
59     - For TypeScript projects:
60```json
61       {
62         "mcpServers": {
63           "my-server": {
64             "command": "node",
65             "args": ["path/to/build/index.js"],
66             "env": {
67               "API_KEY": "key"
68             },
69             "disabled": false,
70             "autoApprove": []
71           }
72         }
73       }
74```
75     - For Python projects:
76```bash
77       # Directly with command line
78       mcp install server.py -v API_KEY=key
79       
80       # Or in settings.json
81       {
82         "mcpServers": {
83           "my-server": {
84             "command": "python",
85             "args": ["server.py"],
86             "env": {
87               "API_KEY": "key"
88             },
89             "disabled": false,
90             "autoApprove": []
91           }
92         }
93       }
94```
95 
96## Step 3: Testing (BLOCKER ⛔️)
97 
98<thinking>
99BEFORE using attempt_completion, I MUST verify:
100□ Have I tested EVERY tool?
101□ Have I confirmed success from the user for each test?
102□ Have I documented the test results?
103 
104If ANY answer is "no", I MUST NOT use attempt_completion.
105</thinking>
106 
1071. Test Each Tool (REQUIRED)
108   □ Test each tool with valid inputs
109   □ Verify output format is correct
110   ⚠️ DO NOT PROCEED UNTIL ALL TOOLS TESTED
111 
112## Step 4: Completion
113❗ STOP AND VERIFY:
114□ Every tool has been tested with valid inputs
115□ Output format is correct for each tool
116 
117Only after ALL tools have been tested can attempt_completion be used.
118 
119## Key Requirements
120- ✓ Must use MCP SDK
121- ✓ Must have comprehensive logging
122- ✓ Must test each tool individually
123- ✓ Must handle errors gracefully
124- ⛔️ NEVER skip testing before completion
125