Step 2: MCP Protocol

Model Context Protocol — the universal adapter for AI tools

1 ExplorePlay below
2 ReadUnderstand
3 BuildHands-on lab
💡 ReflectThink deeper

What Is MCP?

Model Context Protocol (MCP) is an open standard that gives AI agents a structured way to access external tools and data. Think of it as a universal adapter between LLMs and the systems they need to interact with.

ComponentRoleExample
MCP ClientThe AI applicationClaude Desktop, a custom agent
MCP ServerExposes tools via standard protocolCheck Point Management MCP, Jira MCP
ToolsSpecific capabilitieslist_firewalls, get_policy
ResourcesData the server providesFirewall logs, threat intelligence

Why MCP Matters for Security

BenefitRisk
Standardised tool accessEvery MCP connection is a potential data channel
Agents can orchestrate multi-tool workflowsToo many tools = too much blast radius
Tools are discoverable and documentedTool descriptions can be manipulated (poisoning)
Enables security operations automationAutonomous actions without human approval

Available Check Point MCP Servers

The cp-agentic-mcp-playground includes 13 MCP servers covering:

MCP ServerProductRisk Level
Reputation ServiceThreatCloudLow (read-only)
Management LogsSmartLogLow (read-only)
Threat EmulationSandBlastMedium (file submission)
Quantum ManagementSecurity MgmtHigh (rule changes)
Gateway CLIQuantum GWsHigh (CLI commands)
Loading...
Loading...

Think Deeper

An agent has MCP access to both the Reputation Service and Quantum Management. Why is this a risk, and how would you apply least-privilege?

The Reputation Service is read-only (low risk). Quantum Management can modify firewall rules (high risk). If the agent is compromised via prompt injection in a malicious log entry, it could use the Management MCP to open firewall ports. Least-privilege: give investigation agents read-only tools. Only give management access to agents with human-in-the-loop approval.
Key insight: MCP is the new attack surface. Every MCP connection is a capability grant. Least-privilege applies to agents just like it applies to human users — an investigation agent should NOT have gateway management access.

Loading...