Configuring MCP Tools and Resources
1. Enabling MCP Servers in IDE Settings
| Setting | Description | Default Value | Notes |
|---|---|---|---|
| mcp.enabled | Enables or disables MCP server support in the IDE | false |
Set to true to enable MCP integration |
| mcp.serverPath | Path to the MCP server executable or script | null |
Required if MCP is enabled |
| mcp.port | Port number for MCP server communication | 3000 |
Must be available and not in use |
| mcp.timeout | Timeout in milliseconds for MCP server responses | 5000 |
Adjust based on server response times |
Example: Enabling MCP in VS Code settings.json
{
"mcp.enabled": true,
"mcp.serverPath": "/usr/local/bin/mcp-server",
"mcp.port": 3000,
"mcp.timeout": 5000
}
Note: Ensure the MCP server executable has proper permissions and is accessible from your IDE's environment.
2. Selecting Available MCP Tools per Session
| Tool | Description | Availability | Configuration |
|---|---|---|---|
| mcp.tool.name | Name of the MCP tool to be used in session | NEW | "toolName" |
| mcp.tool.version | Version of the MCP tool to use | NEW | "v1.0" |
| mcp.tool.enabled | Boolean flag to enable or disable tool in session | NEW | true or false |
Example: Enabling a Tool in Session
{
"mcp": {
"tools": [
{
"name": "CodeAssistant",
"version": "v2.0",
"enabled": true
}
]
}
}
Note: Tool selection is session-specific and does not persist across sessions.
3. Restricting MCP Tool Permissions
| Permission | Description | Restriction Level | Notes |
|---|---|---|---|
| mcp.tool.execute | Execute a tool command | DEPRECATED | Replaced by specific permission scopes |
| mcp.tool.read | Read tool metadata and configuration | Standard | Required for tool discovery |
| mcp.tool.write | Modify tool configuration and metadata | High | Can alter tool behavior or access |
| mcp.tool.delete | Delete tools from the system | Very High | Can remove critical tools |
| mcp.tool.invoke | Invoke a tool with specific parameters | Standard | Used for executing tools in workflows |
Example: Setting Tool Permissions in MCP Config
{
"tools": [
{
"name": "data-analyzer",
"permissions": [
"mcp.tool.read",
"mcp.tool.invoke"
]
}
]
}
Note: Permissions can be scoped to specific tools or groups of tools. Use
mcp.tool.* for broad access control.
Warning: Granting
mcp.tool.write or mcp.tool.delete permissions to untrusted users can lead to system compromise.
4. Managing Multiple MCP Server Connections
| Connection | Configuration | Status | Notes |
|---|---|---|---|
| server-1 | { "url": "http://localhost:3000", "auth": "token123" } |
ACTIVE | Primary MCP server for code generation |
| server-2 | { "url": "https://mcp.example.com", "auth": "bearer abcdef" } |
INACTIVE | Legacy server, deprecated since v2.0 |
| server-3 | { "url": "wss://ws.mcp.dev", "auth": "oauth2" } |
BETA | Experimental server for AI-assisted debugging |
Example: Connecting to Multiple Servers
const mcpClient = new MCPClient({
servers: [
{ id: 'server-1', url: 'http://localhost:3000', auth: 'token123' },
{ id: 'server-2', url: 'https://mcp.example.com', auth: 'bearer abcdef' },
{ id: 'server-3', url: 'wss://ws.mcp.dev', auth: 'oauth2' }
]
});Workflow: Switching Between Servers
- Initialize MCP client with multiple server configurations
- Set default server using
setDefaultServer('server-1') - Query specific server using
queryServer('server-3', 'debug') - Monitor connection status via
getConnectionStatus('server-2')
Server Management Summary
- Primary server is used for default operations
- Legacy servers can be kept for migration paths
- Beta servers are for experimental features only
5. Debugging MCP Tool Call Failures
| Issue | Description | Solution |
|---|---|---|
| Invalid Tool Response | Tool returns malformed or unexpected data | Validate response schema before processing. Log raw responses for inspection. |
| Timeout Errors | Tool takes longer than allowed to respond | Increase timeout limits or optimize tool logic. Use async handling. |
| Authentication Failures | Tool access denied due to invalid credentials | Verify token validity and permissions. Refresh tokens if needed. |
| Network Errors | Connection issues during tool invocation | Implement retry logic with exponential backoff. Check network connectivity. |
Example: Logging Tool Call Errors
const toolCall = async (toolName, params) => {
try {
const response = await fetch(`/api/tools/${toolName}`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify(params)
});
if (!response.ok) {
throw new Error(`HTTP error! status: ${response.status}`);
}
return await response.json();
} catch (error) {
console.error(`Tool call failed for ${toolName}:`, error);
throw error;
}
};
Note: Use structured logging to capture tool call metadata including input parameters, response times, and error codes for effective debugging.
6. Monitoring MCP Server Logs
| Log Level | Description | Usage |
|---|---|---|
| trace | Most detailed logging, includes all internal operations and data flow | For debugging complex issues or performance bottlenecks |
| debug | Includes debug information, such as internal state changes and method calls | Use during development or troubleshooting specific components |
| info | General operational messages, including startup and shutdown events | Standard monitoring for production environments |
| warn | Warning messages indicating potential issues that don't stop execution | Alerts for non-critical conditions or deprecated features |
| error | Error messages indicating failures or exceptions that halt execution | Immediate attention required for system stability |
Example: Configuring Log Level in MCP Server
{
"logging": {
"level": "info",
"file": "/var/log/mcp-server.log"
}
}
Note: Log levels can be adjusted dynamically without restarting the server.
Warning: Enabling
trace logging in production can significantly impact performance and disk space usage.
Log Level Summary
- trace: For in-depth diagnostics and debugging
- debug: For development and component-specific issues
- info: Standard operational visibility
- warn: Non-critical issues that require attention
- error: Critical failures requiring immediate action
7. Versioning MCP Server Configurations
| Configuration Element | Description | Versioning Strategy | Notes |
|---|---|---|---|
| mcp.config.version | Version identifier for the MCP server configuration | NEW in v1.2 | Used to track configuration changes and ensure compatibility |
| mcp.config.schemaVersion | Schema version of the configuration file | Incremented on schema changes | Ensures backward compatibility and migration paths |
| mcp.config.metadata | Metadata section for version control and tracking | NEW in v1.2 | Includes timestamps, author, and change logs |
Example: Versioned Configuration File
{
"version": "1.2.0",
"schemaVersion": "1.2",
"metadata": {
"author": "dev-team",
"lastModified": "2026-04-05T10:00:00Z",
"changeLog": [
{
"version": "1.2.0",
"description": "Updated server config for new toolset"
}
]
},
"server": {
"port": 3000,
"host": "localhost"
}
}
Note: Versioning ensures that configuration changes can be tracked, rolled back if necessary, and validated for compatibility with the MCP server version.
8. Disabling Unused MCP Servers
| Server ID | Status | Notes |
|---|---|---|
| mcp-server-1 | Enabled | Default server for code completion |
| mcp-server-2 | Disabled | Unused for current project |
Example: Disable MCP Server via Configuration
{
"mcp": {
"servers": {
"mcp-server-1": {
"enabled": true
},
"mcp-server-2": {
"enabled": false,
"reason": "Not required for current project"
}
}
}
}
Note: Disabling unused servers reduces resource consumption and improves performance.