Quick Start

This guide provides two shortest onboarding paths, with a clear recommended order.

The default recommended path is the Skill path, because it best matches the core purpose of MCP2Skill: converting MCP tools into Skills that load on demand, reducing Token consumption, and relieving you from maintaining configurations back and forth across different AI clients.

Before You Start

Before getting started, we recommend confirming:

• You already know where the MCP services you want to connect come from
• If you plan to import an old configuration, the original application's configuration file still exists
• If you plan to add remote services manually, you have the correct URL, request headers, or authorization requirements ready

Choose a Primary Path First

Recommended Path: Skill Path

Suitable if you want to:

• Convert MCP tools into Skills
• Serve AI Agents
• Use on-demand loading to further save Tokens and reduce repeated context

Recommended flow:

1. Import or add MCP services
2. Confirm the services and tools are available
3. Generate Skills
4. Install them to an AI Agent
5. Verify the benefits in a real AI workflow

Secondary Path: Gateway Path

Suitable if you want to:

• Reuse the same set of MCP configurations across multiple AI clients
• Avoid configuring MCP repeatedly in every client
• Replace each client launching MCP on its own with a single unified gateway

Recommended flow:

1. Import or add MCP services
2. Confirm the services and tools are available
3. Create a workspace
4. Copy the endpoint or JSON configuration
5. Connect it in an external AI client

Step 1: Import an Existing Configuration

This is the most time-saving approach.

1. Open MCP in the left navigation
2. Click Import in the top-right corner
3. Choose the import source
4. Preview the list of detected services
5. Check the services you want to import
6. Run the import

Once the import completes, you will immediately see these services in the MCP service list.

If You Have No Existing Configuration: Add MCP Services Manually

If you have no existing configuration, you can create services directly.

1. Go to MCP
2. Click Add in the top-right corner
3. Choose the service type
4. Fill in the basic information and connection configuration
5. Enable the service if needed
6. Save the configuration

We recommend adding and verifying at least one service first to confirm the whole flow works before organizing services in bulk.

Step 2: Start and Confirm the Service Is Available

After adding or importing, first confirm the service itself works.

1. Open the target service in the MCP service list
2. Check whether its status is connected or running
3. Open the tool list in the detail page
4. Confirm the tool names have loaded successfully

If it is a remote service that requires authorization, complete the authorization flow first, then return to the tool list to confirm the result. Whether you take the Skill path or the gateway path afterwards, this step cannot be skipped.

Recommended Path: Generate and Install Your First Skill

If your goal is to save Tokens, do not rush to create a workspace next; instead, convert the available MCP tools into Skills first.

1. Open the detail page of an MCP service that is already available, or organize a workspace first
2. Click Generate Skill
3. Preview the generated result
4. Choose to export to a directory, or install directly to an AI Agent
5. Verify in a real AI Agent

We recommend using "one high-frequency tool service" for your first round of Skill verification, as this makes the benefits easiest to perceive.

Secondary Path: Create Your First Workspace

If your goal is to connect multiple AI clients through a single entry point, you can create a workspace once at least one service is available.

1. Click the add button in the workspace area on the left
2. Fill in the workspace name, display name, and description
3. Check the MCP services to include in this workspace
4. If needed, further disable some tools within the workspace
5. Save the workspace

After the workspace is saved, a gateway endpoint that external clients can use is generated automatically.

Secondary Path: Copy the Endpoint or JSON Configuration

The next step is to connect MCP2Skill to your external AI applications.

1. Open the workspace detail page
2. Locate the gateway endpoint area
3. Click the copy menu
4. Choose to copy the endpoint address, or copy the JSON configuration

If you just want a quick test first, you can also use the endpoint of the system workspace ALL directly.

Verify the Results

If you took the Skill path, we recommend verifying:

1. Whether the Skill has appeared in the target AI Agent directory
2. Whether the AI Agent can correctly recognize and use the Skill
3. Whether you achieved the reuse and Token control benefits you expected

If you took the gateway path, we recommend verifying:

1. Restart or refresh the MCP configuration in the external AI application
2. Send a request that will use an MCP tool
3. Return to the MCP2Skill dashboard, statistics, or logs page to observe the results

As long as you can see the call records, the pipeline is working end to end.

Recommended Order Summary

• If your goal is saving Tokens: take the Skill path first
• If your goal is unified access for multiple AI clients: take the gateway path first
• The Skill path addresses Tokens and workflow efficiency
• The gateway path addresses integration compatibility and unified management
• The two do not conflict; many users organize MCP first and then use both paths at the same time

If You Run Into Problems Along the Way

• Service will not start: see MCP Service Management
• Skill installation or recognition is wrong: see Generate and Install Skills
• Remote authorization fails: see Remote Services and OAuth Authorization
• Client cannot connect: see Connect External AI Clients
• No tools are showing: see FAQ and Troubleshooting