MCP Services¶

MCP Services are used to allow Agents to connect to external tools, systems, or services. They can provide configured tool capabilities to Agents for use, such as querying data, invoking automation tools, accessing internal systems, or performing specific operations.

Unlike Skills, MCP Services are closer to "tool gateways". If an MCP Service connects to a real system, special attention must be paid to permissions, data scope, and operational risks.

Configuration Logic¶

Obsy Agent Teams manage MCP Services using a "global configuration first, then enable for Agent" approach:

• Global Configuration: Administrators first configure the MCP Services available to the team in the Settings Center.
• Agent Enablement: Enter a specific Agent's workspace and enable only the MCP Services that the Agent needs to access.

This approach allows tool capabilities to remain centrally managed while preventing unrelated Agents from accessing sensitive systems. For example, a production troubleshooting Agent can enable read-only query MCP Services, while a regular document organization Agent does not need to enable production system-related MCP Services.

Usage Recommendations¶

Before configuring an MCP Service, it is recommended to confirm:

• Whether the MCP Service connects to a sensitive system.
• Whether the Agent only needs read-only capabilities.
• Whether write operations require approval.
• Whether only specific Agents should use it.
• Whether MCP Service failures will affect task execution or external systems.

For scenarios involving production changes, resource deletion, permission modifications, deployment rollbacks, etc., it is recommended to retain manual confirmation or approval mechanisms.