Workspace Basics

If an MCP service is the raw material, then a workspace is the organized toolbox. It determines which set of tools you want external AI applications to see.

Why Use Workspaces

Workspaces are best suited for problems like these:

• Different projects need different sets of tools
• You don't want to expose every tool from a given MCP service
• You want to expose different endpoints to different clients

Compared with handing every service directly to external clients, workspaces make it much easier to control scope.

What a Workspace Contains

A workspace typically includes:

• Name identifier
• Display name
• Icon
• Description
• Selected MCP services
• Tools disabled within the workspace
• Automatically generated workspace endpoint
• Skill generation mode

Creating a Workspace

The typical steps are:

1. Click the add button in the workspace area on the left
2. Fill in the name identifier and display name
3. Set the icon and description as needed
4. Select the MCP services to include
5. Choose the Skill generation mode
6. Further filter the tools
7. Save

After saving, the workspace appears immediately in the list on the left.

Skill Generation Mode

A workspace controls the structure of the output when Skills are generated from it. This option is set when creating or editing the workspace.

Aggregated Skill Generation Off

By default, Aggregated Skill Generation is off. In this state, exporting or installing Skills from the workspace generates an independent Skill for each MCP service in the workspace.

This mode is suitable when:

• You want the AI Agent to load capabilities on demand more precisely, per specific MCP capability
• The workspace contains multiple MCP services with very different purposes
• You don't want a single Skill's description to become too broad

These independent Skills still call the workspace's endpoint, so workspace-level tool filtering remains in effect.

Aggregated Skill Generation On

With Aggregated Skill Generation enabled, generating Skills from the workspace produces a single aggregated Skill that represents the entire workspace.

This mode is suitable when:

• You want to hand a group of MCP services to the AI Agent as one complete toolbox
• The workspace itself corresponds to a well-defined task scenario
• You want to write a unified Skill description for the entire workspace

Only when aggregated generation is enabled does the workspace form show the Skill Description input field and the auto-generate button. When aggregated generation is off, MCP2Skill reuses each MCP service's own Skill description.

Only Enabled Services Belong in a Workspace

The value of a workspace rests on its services already being usable. In practice, follow this order:

1. Configure the MCP service itself first
2. Confirm the service is running and its tool list looks normal
3. Then add it to a workspace

If a service is later disabled, the corresponding tools in the workspace stop working as well.

The System Workspace ALL

The system workspace ALL is an aggregated workspace automatically maintained by MCP2Skill. It typically has these characteristics:

• Automatically aggregates the tools of all currently enabled MCP services
• Good for quickly verifying overall connectivity
• Not suitable as a long-term solution for fine-grained access control
• Some editing capabilities differ from regular workspaces

If you're just getting started, testing with ALL is usually convenient; once you begin using MCP2Skill regularly, it's better to create your own business workspaces.

Editing a Workspace

As requirements change, you can edit a workspace to:

• Adjust the display name or description
• Change the icon
• Switch the Skill generation mode
• Add or remove MCP services
• Re-filter tools within the workspace scope

Treat a workspace as "the list of tools available to external clients," rather than a simple grouping for notes.

Check Before Deleting a Workspace

Before deleting, it's worth checking:

• Whether any external clients are still using the workspace endpoint
• Whether you only want to temporarily disable some tools
• Whether you can keep the workspace and just adjust its services and tool scope

If the workspace is only temporarily unneeded, narrowing its contents first is usually safer than deleting it outright.

Naming Conventions That Work Well

Workspace names should reflect purpose, not technical details. For example:

• web-dev
• data-analysis
• writing
• client-a

This makes endpoints easier to identify when copying them later, and makes it easier to distinguish them across external AI applications.

Related Chapters

• To further filter tools and copy endpoints: see Tool Selection and Endpoints
• To configure the services themselves first: see MCP Service Management
• To connect the workspace to external clients: see Connect External AI Clients