Connectors (MCP)

Connectors are LanMate’s standardized channel for calling external tools and services, implemented on the MCP (Model Context Protocol) protocol. By configuring MCP services, LanMate can access the local file system, call third-party APIs, run database queries, and more during conversations — extending the AI assistant’s reach to the tools you use every day.

This article covers the basics of the MCP protocol, how to add and manage MCP services, and things to note when using them.

What Is a Connector

MCP is an open protocol that standardizes communication between an AI assistant and external tools. Each MCP service exposes a set of “tools”; once enabled, the AI assistant can call these tools on demand in conversations to get tasks done.

LanMate supports two transport modes, stdio and http, corresponding to local and remote services respectively:

Which mode to choose depends on how the service itself is provided: use stdio when the provider gives you an executable command, and http when they give you an accessible URL.

Adding an MCP Service

Go to MCP configuration on the Settings page to view all currently configured services, and to add, edit, or delete them.

Before adding, get the service’s connection information from its provider: for a local service, that’s usually the launch command and required arguments; for a remote service, the access URL and authentication method.

Steps to add a new service:

1. Open the Settings page and go to MCP configuration.
2. Click Add Service, fill in the service name (required), and optionally add a description so you can recognize the service’s purpose later.
3. Choose a transport type (stdio or http).
4. Fill in the configuration for the chosen mode:
   • stdio mode: Fill in the launch command, arguments, and environment variables. The command starts the service process locally; environment variables can carry authentication information such as API Keys.
   • http mode: Fill in the service URL and HTTP headers. Authentication information is passed through HTTP headers.
5. Adjust the timeout as needed (default 30 seconds; usually no change required).
6. Save the configuration. The system automatically tries to connect to the service; once connected, it’s available in conversations.

Besides filling in the form, the configuration interface also supports a JSON mode for users who prefer editing the configuration text directly.

When adding or editing a service, the system automatically tests whether the connection works and lets you see which tools the service provides, so you can confirm the configuration is correct. Even if the connection test fails, you can save the configuration first and fix it later via Edit after troubleshooting.

Managing Existing Services

In the MCP configuration list, each service shows its name and connection status (connecting, connected, error), so you can quickly tell whether each service is usable.

Added MCP services support the following management actions:

• Enable/Disable: Turn a configured service on or off anytime. When disabled, the service configuration is kept but the AI assistant no longer calls its tools; re-enable it to restore.
• Edit: Modify the service’s configuration (command, URL, environment variables, etc.); the system automatically reconnects after saving.
• Delete: Remove services you no longer need. Deletion requires a second confirmation; once confirmed, all of the service’s configuration is cleared and cannot be recovered — proceed with caution.

When a service connection has an error, it’s usually related to an invalid command, a changed URL, or expired credentials. Edit the configuration and save — it reconnects automatically.

Using in Conversations

Once an MCP service is enabled, the AI assistant can call the tools it provides in conversations:

1. In MCP configuration, confirm the target service is enabled and connected.
2. Go back to the conversation and describe what you need directly to the AI assistant.
3. With your authorization, the AI assistant calls the tools provided by enabled services to complete the task.

The AI assistant only calls tools from enabled services — services that aren’t enabled or configured are never accessed. If a task needs a capability that’s temporarily missing, enable the corresponding service in MCP configuration first, then return to the conversation and keep asking; when you no longer need it, just disable it in the configuration — no need to delete the service.

Notes and Security Advice

When using MCP services, please note the following:

• Local configuration storage: MCP service configurations (commands, URLs, environment variables, etc.) are stored locally.
• Enable with care: The access scope of a custom MCP service is determined by your configuration — LanMate doesn’t preset a scope, so evaluate carefully before enabling.
• Trusted sources: When adding a service, make sure the command or URL comes from a trusted source. Don’t add services from unknown origins.
• Protect sensitive information: When environment variables or HTTP headers contain sensitive information (such as API Keys), keep them well protected and avoid disclosure.
• Clean up promptly: For services you no longer use, disabling or deleting them promptly is recommended to reduce unnecessary exposure.

If you run into problems while configuring, first check that the command, URL, and credentials are entered correctly, then fix them via Edit and retry the connection.