Upload MCP Event Logs
Upload a JSON file containing MCP events for batch processing. The system will automatically analyze each event for security threats and send alerts for any suspicious activity.
Expected File Format
Your JSON file should contain MCP event data in this exact format:
Single Event Example:
{
"timestamp": "2023-12-01T10:30:00Z",
"user": "user@example.com",
"agent": "claude-3",
"tool": "filesystem",
"prompt": "Read the configuration file",
"response": "File contents: {...}"
}
Multiple Events Example:
[
{
"timestamp": "2023-12-01T10:30:00Z",
"user": "user@example.com",
"agent": "claude-3",
"tool": "filesystem",
"prompt": "Read the configuration file",
"response": "File contents: config.json"
},
{
"timestamp": "2023-12-01T10:31:00Z",
"user": "user@example.com",
"agent": "claude-3",
"tool": "operation",
"prompt": "Read the configuration file",
"response": "Operation: read\nFile: /etc/config.json"
}
]
Supported formats: Single JSON object or array of objects
Real-time API Ingestion
For real-time event ingestion, use the API endpoint:
POST /api/events
Single Event:
{
"timestamp": "2024-01-15T14:30:00Z",
"user": "alice",
"agent": "ChatBot",
"tool": "web_search",
"prompt": "Search for recent security vulnerabilities",
"response": "Found 15 recent CVE entries..."
}
Multiple Events:
[
{
"timestamp": "2024-01-15T14:30:00Z",
"user": "alice",
"agent": "ChatBot",
"tool": "web_search",
"prompt": "Search for data",
"response": "Results found"
},
{
"timestamp": "2024-01-15T14:31:00Z",
"user": "bob",
"agent": "AssistantBot",
"tool": "os_shell",
"prompt": "Run whoami command",
"response": "user123"
}
]
cURL Example:
curl -X POST https://mcpsentinel.com/api/events \
-H "Content-Type: application/json" \
-d '{
"timestamp": "2024-01-15T14:30:00Z",
"user": "alice",
"agent": "ChatBot",
"tool": "web_search",
"prompt": "Search for information",
"response": "Search completed successfully"
}'
Security Detection Rules
Events are automatically flagged based on the following criteria:
Dangerous Tools
os_shell
- Direct shell access
Suspicious Keywords
- • "delete all"
- • "ignore instructions"
- • "export database"
- • "shell", "sudo"
- • "<script>", "<system>"
Size Limits
- • Prompts > 3,000 characters
- • Responses > 5,000 characters
Injection Patterns
- • SQL injection attempts
- • Command injection patterns
- • Script injection attempts