MCP Service Management

The MCP page is the foundation of all of MCP2Skill. Every workspace, endpoint, and Skill is built on the premise that your MCP services are configured correctly.

What You Can Do on the MCP Page

In the current version, you can:

• View all MCP services
• Search services
• Add services
• Discover and install services from the MCP Market
• Import configurations
• Export configurations
• Enable or disable services
• Open the detail page to view tools, configuration, and call history
• Generate a Skill from a single service

Supported Service Types

STDIO

Best suited for services launched by local commands, commonly seen with MCP services run from the command line.

Fields you typically need to pay attention to include:

• Command
• Arguments
• Environment variables
• Working directory

SSE

Best suited for MCP services connected via a remote SSE address. Some services may also require OAuth authorization.

Common fields:

• Service URL
• Custom request headers

Streamable HTTP

Best suited for remote HTTP endpoint MCP services.

Common fields:

• Service URL
• Custom request headers

Discovering Services from the MCP Market

In the current source code, the MCP page includes a Discover entry for browsing built-in marketplace data and quickly installing common MCP services.

When using the marketplace entry, it is recommended to review the service details and required environment variables before deciding to install. After installation, you should still return to the service list or detail page to confirm the connection status, tool count, and authorization state.

Marketplace installation is ideal for quickly onboarding common services; if you have custom commands, local scripts, or private remote services, you can still add them manually.

Adding a Service

When adding a new MCP service, it is recommended to fill in the fields in the following order:

1. Select the service type
2. Enter the name identifier
3. Enter the display name
4. Add a description
5. Fill in the connection configuration based on the type
6. Enable the service if appropriate
7. Save

The name identifier is typically used for the tool namespace and endpoint identifier. It is recommended to keep it stable and avoid changing it frequently.

Editing a Service

Edit mode is mainly suited for handling the following situations:

• Changing the display name or description
• Updating the URL, command arguments, request headers, or environment variables
• Adjusting the working directory
• Fixing incorrect configuration

If you find that a service clearly exists but never connects, first go back to the edit page and check whether the configuration is complete.

Enabling and Disabling

Each service can be enabled or disabled individually.

• When enabled, MCP2Skill will attempt to establish a connection and fetch the tool list
• When disabled, the service will no longer expose callable tools
• Related tools in workspaces will become unavailable accordingly

After importing a large number of services, it is recommended to check your key services one by one to confirm they are running successfully.

Understanding Service Statuses

Common statuses can be understood as follows:

For remote services, you may also see authorization-related statuses. See Remote Services and OAuth Authorization for details.

Viewing the Tool List

The tool list on the service detail page is very important — it helps you determine whether the service is actually usable.

Focus on three things:

• Whether the number of tools matches your expectations
• Whether tool names display correctly
• Whether any tools have been manually turned off

If the service status is healthy but the tool list is empty, it usually means initialization did not complete successfully, or the remote service has not been authorized yet.

Enabling and Disabling Tools at the Service Level

The current version supports toggling individual tools within a single service.

This is suitable for scenarios such as:

• A tool is high-risk and you do not want to expose it to all workspaces
• A tool is still in a testing phase and should be turned off for now
• A service has too many tools and you want to narrow the exposed scope first

Tools disabled at the service level will not become available again in workspaces.

Viewing Configuration Details and Call History

Besides the tool list, the service detail page also shows:

• Configuration details
• Call history
• Gateway endpoints
• A JSON configuration copy entry

This means you usually do not need to switch back and forth between pages to complete the full cycle of "check the configuration, copy the endpoint, and review call results."

What to Check Before Deleting a Service

Deletion is irreversible. Before deleting, it is recommended to confirm:

• Whether the service is still depended on by any workspace
• Whether any external AI applications are still using its endpoint
• Whether you still need its historical call records for reference

If you simply do not need it for now, disabling it first is usually safer than deleting it outright.

Related Chapters

• Remote service authorization issues: see Remote Services and OAuth Authorization
• Bulk import or export: see Import and Export Configuration
• Organizing services into workspaces: see Workspace Basics