> For the complete documentation index, see [llms.txt](https://docs.daylite.app/help/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.daylite.app/help/mac-fundamentals/integrations/local-mcp-server.md). # Local MCP Server Learn about what Daylite's Local MCP server is, how to turn it on safely, and how to connect Claude Desktop or Codex so an AI assistant can find, create, and update your Daylite records. {% hint style="warning" icon="eye-slash" %} Daylite’s Local MCP server is in Private Preview. Features and approaches are still evolving, and you may encounter a few bugs. Your feedback will help shape what we build next and how we build it. Please share feedback on the Local MCP server or this document on the [Daylite Local MCP community](https://community.daylite.app) page. We’d also love to see interesting prompts you’ve tried and the results you’ve had. {% endhint %} ## What is a Local MCP server? A Local MCP (Model Context Protocol) server gives an AI assistant a set of tools it can use with Daylite. Each tool does one job — for example, finding a person or creating a task. Each tool has set inputs and returns data in a set format. Together, these tools control what the AI can and cannot do with your Daylite data. "Local" means the server runs on your Mac. Daylite and your AI app must be open on the same Mac, and nothing on your network or the internet can connect to the server. The server is off by default, and it shuts off when you quit Daylite. The Local MCP server lets an AI assistant find, view, create, update, and connect records. It cannot delete records, and each search can return up to 100 records at a time. More tools and features are being added. See the Daylite Local MCP Tool Reference for the full list. ## Before You Begin AI can be useful when it is used for the right work. It will not replace your sales team or run your business for you, but it can help people work faster by drafting, finding, summarizing, and acting on Daylite data. You still need to use your judgment. You decide what is correct, what is right for the task, and what is worth acting on. The AI model reads each request and decides what actions to take, so results can vary — the same request may not always give the same result. Check the output before you rely on it, especially when the AI writes data back to Daylite. ### Settings & Requirements Before you start, make sure you have: * macOS Sequoia 15.2 or later * [Latest Daylite Beta](/help/center/install-uninstall-and-update/update-settings.md#change-update-channel) * A supported AI app installed — such as [Claude Desktop](https://claude.com/download) or the [ChatGPT app](https://chatgpt.com/download) for Mac, which runs Codex The Local MCP server is in Daylite under **Settings** > **MCP Server**, and it is off by default. Turn it on there, then connect your AI app by following the setup guide for that app. ### Securing Your Data Daylite's Local MCP server has tools that can read or change Daylite data. When the AI model picks a tool, the AI app runs it through its MCP host. The app may then send the result to its AI provider so the model can keep working on your request. Some Daylite data may leave your Mac when you use the MCP server. If your AI app uses a cloud service, it will usually send tool results to the AI provider. How that data is stored, cached, logged, or kept depends on the app, provider, account type, and settings. The server cannot delete records, **but it can overwrite fields in a record.** Your AI app may ask you to approve each Daylite tool call, and some accounts or company settings may let you approve future calls by default. Daylite does not have an audit log for MCP changes, so it is best to ask the AI to show you its planned changes first and review them before you approve the action. Daylite always follows its own permissions. If your Daylite user can read a record, the AI can read it; if your user can edit a record, the AI can edit it. To block access to a record, change the permissions for your Daylite user. As with any integration, you should review whether your setup meets your compliance needs. For details about GDPR and other privacy rules, check your AI provider's privacy and data policies. ## Using the Local MCP Server Once you meet the requirements, turn on the server in Daylite and connect a supported AI app. Daylite currently gives setup steps and support for two apps. ### Claude Desktop Claude Desktop has the easiest setup. Start the server in Daylite, then use **Configure Claude** to install the Daylite connection for you. {% content-ref url="/pages/fmQBURjCnJjlu87TTTuW" %} [How to Connect Daylite to Claude Desktop using MCP](/help/mac-fundamentals/integrations/local-mcp-server/how-to-connect-daylite-to-claude-desktop-using-mcp.md) {% endcontent-ref %} ### Codex Codex runs inside the ChatGPT app for Mac. There is no one-click setup — you must add Daylite as a custom MCP server in the app settings. Regular ChatGPT chat cannot connect to MCP servers, so this works only in Codex. {% content-ref url="/pages/a6RoQPb2gBg984HoK0t9" %} [How to Connect Daylite to Codex using MCP](/help/mac-fundamentals/integrations/local-mcp-server/how-to-connect-daylite-to-codex-using-mcp.md) {% endcontent-ref %} ### Daylite Local MCP Tool Reference The tools let an AI assistant find, view, create, update, and connect Daylite records. Deleting is not supported, searches can return up to 100 records at a time, and more tools are being added. For a full list of what each tool can do, see the Daylite Local MCP Tool Reference. It shows which fields you can search, read, create, and update for each record type, and it covers linking rules, statuses, priorities, and planned features. {% content-ref url="/pages/3c5pHCi5h2Aeb8KIL2Vh" %} [Daylite Local MCP Tool Reference](/help/mac-fundamentals/integrations/local-mcp-server/daylite-local-mcp-tool-reference.md) {% endcontent-ref %} ## Frequently Asked Questions **Which AI applications can I use with the MCP server?** Daylite currently tests, documents, and supports **Claude Desktop** and **Codex** in the ChatGPT app for Mac. Other apps that support MCP may also connect to Daylite's MCP server, but Daylite does not officially support them. **What actions does the MCP server support?** The server has tools that can read and write data. Each action is limited by the permissions of the current Daylite user, and deleting records is not supported. **How do I limit who can use the MCP server?** Daylite does not yet have a separate per-user setting for MCP. The current Daylite user's record permissions still control what the AI can read or change. **How can I prevent an AI model from accessing certain data in my database?** Daylite permissions are enforced. If your Daylite user can read a record, the AI can read it; if your user can edit a record, the AI can edit it. To block access, change the permissions for your Daylite user. **How do we prevent deleting anything with the MCP server?** Deleting is not supported, so the server cannot delete records. **I don't want an AI application to connect to my data. How do I prevent this?** The Daylite MCP server is off by default. To use it, you must turn it on and connect your AI app, and you can turn the server off at any time. **Is there an audit log of what the AI creates or changes?** Daylite does not currently have a dedicated MCP activity audit log. **Is my data stored or cached anywhere, or does it just pass through?** Data from the MCP server is sent to your MCP app. It may also be kept in chat history, cached for a short time, or added to operations and safety logs. This depends on the AI app, your settings, your account type, and the AI provider's data retention rules. **Is this GDPR compliant?** As with any integration, you should decide if your setup meets your compliance needs. For details about GDPR and related rules, review your AI provider's privacy and data policies. **Is there an HTTP endpoint?** No. Daylite uses local stdio transport and does not provide a Streamable HTTP endpoint. Daylite must be open on the Mac for the integration to work. **Does this work on iPhone or iPad?** No. Daylite's Local MCP server runs on your Mac, and Daylite and your AI app must both be running on that same Mac. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.daylite.app/help/mac-fundamentals/integrations/local-mcp-server.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.