> ## Documentation Index
> Fetch the complete documentation index at: https://docs.llmeasy.ru/llms.txt
> Use this file to discover all available pages before exploring further.

# How to configure MCP in Claude Desktop

> Understand local MCP, remote MCP, connection status, and the difference between MCP and model APIs in Claude Desktop.

## Short answer

Claude Desktop can connect to local tools and data sources through MCP servers. You declare the server in Claude Desktop's MCP configuration, restart or refresh the connection, and check the connection status in the app.

MCP is not a model API. It lets Claude Desktop access tools or context. Model requests still depend on the model and account path currently used by Claude Desktop.

## When you need it

* You want Claude Desktop to access local files, databases, or development tools
* You see an MCP server config but do not know where it belongs
* You are comparing local MCP and remote MCP
* You think MCP can replace API Key / Base URL
* You need to check whether MCP is connected

## Local MCP vs remote MCP

| Type              | Meaning                                                                                  |
| ----------------- | ---------------------------------------------------------------------------------------- |
| Local MCP         | Runs a server on your machine, often for local tools, files, or development environments |
| Remote MCP        | Connects to a remote service, often for cloud tools or team services                     |
| Desktop extension | A more desktop-oriented installation and distribution format                             |

## Recommended setup

1. Confirm that you are connecting an MCP server, not a model provider.
2. Add the local or remote MCP config according to the Claude Desktop docs.
3. Restart or refresh Claude Desktop.
4. Check the MCP connection status in Claude Desktop.
5. Give the MCP server only the permissions and environment variables it needs.

## Common mistakes

* Treating an MCP server as a Base URL.
* Putting a model API Key into MCP configuration.
* Assuming MCP changes model cost.
* Forgetting that an MCP server may access sensitive local data.
* Checking only the Claude Desktop UI when the local server is not running.

## About LLMEasy

LLMEasy handles model API access, not MCP servers. Claude Desktop MCP connects tools. LLMEasy-style API layers handle model requests, balance, and usage.

If you need both tool access and unified model routing, configure MCP and API access separately.

## Related docs

* [MCP vs API Key and Base URL](/en/faq/concepts/mcp-vs-api-key-base-url)
* [Claude Code vs Claude Desktop](/en/faq/claude-code/claude-code-vs-claude-desktop)
* [Does Claude Code use a Claude.ai account or an API Key?](/en/faq/claude-code/claude-account-or-api-key)
* [OpenAI-compatible API vs Anthropic-compatible API](/en/faq/concepts/openai-compatible-vs-anthropic-compatible)

## References

* [Getting started with local MCP servers on Claude Desktop](https://support.anthropic.com/en/articles/10949351-getting-started-with-local-mcp-servers-on-claude-desktop)
* [Getting started with custom connectors using remote MCP](https://support.anthropic.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp)