LMCP (local-mcp-releases) is used to connect MCP-compatible AI assistants to applications and accounts on the user's computer. It gathers relevant information from email, messages, calendars, and files through the Model Context Protocol, and with the help of available tools turns this context into actual operations, reducing the need to search back and forth between applications and copy and paste.
In terms of use cases, LMCP is aimed at scenarios where users want direct access to local or already synced data within an AI assistant. The README gives an example: an email, a message, and a document may each contain only part of the answer, and within the access scope authorized by the user, the assistant can find these fragments across applications and accounts and combine them. The text emphasizes that these are only examples of workflows the assistant can build based on available tools, not guaranteed autonomous results; sources must be accessible, and the assistant's capabilities and the permissions of each tool still apply.
In terms of main capabilities, the material states that it provides 201+ local tools, covering integrations for email, calendars, contacts, messages, tasks, files, documents, browser automation, and more, with examples including Apple Mail, Outlook, Microsoft Teams, Slack, WhatsApp, Signal, Google Drive, OneDrive, Microsoft 365, Notes, Reminders, Microsoft To Do, OmniFocus, Notion, and ServiceNow. Not all integrations are available on both operating systems: Apple native tools require macOS; connected services may require separate login or authorization; some tools use local caches or synced data rather than full cloud accounts. The README reminds users to rely on the current information in the tool reference and guide, rather than a fixed count in the README.
In terms of usage, you first need to download and install the LMCP application (this must be done first, because the tools run on the user's computer). Installation paths include: the Mac signed installer package (.pkg) or the Windows installer download page; the macOS terminal command curl -fsSL 'https://www.local-mcp.com/install?ref=github-releases' | bash; the Windows PowerShell command irm https://www.local-mcp.com/install-windows?ref=github-releases | iex (installs per user, no administrator privileges or Node.js required); the Claude Desktop extension can install .mcpb with one click via www.local-mcp.com/install/mcpb; the npm method (requires Node.js) uses npx -y local-mcp@latest setup to configure detected compatible clients. After installation, the AI client must be fully restarted because MCP tools are loaded at startup.
The repository can also be used as a Cursor plugin, which adds the local-mcp MCP server, a rule telling the Cursor agent when to use LMCP, and skills for email, calendars, messages, files, and Office; before use, the LMCP application must also be installed first, and the plugin will start the LMCP server on the computer.
Local desktop/CLI clients connect differently from web assistants. The installer configures compatible local clients it detects. Web assistants (ChatGPT, Claude.ai, Grok, Perplexity) connect through https://www.local-mcp.com/mcp As a custom connector using OAuth, you need to install the local app and enable Cloud Data Forwarding in LMCP (off by default); if forwarding is disabled, the connector only exposes setup_install. The README points out that having an endpoint alone does not equal a verified connection; you need to complete authorization and confirm that the tools are available in the client.
Regarding applicable scenarios and limitations, locally available and read-only operations may work offline; cloud models, connected services, sending, and remote updates may require a network connection. Local execution does not mean the cloud AI will not receive data: the assistant or connected service chosen by the user may still receive what is needed for the request, including when the desktop client uses a cloud model. Cloud Data Forwarding is optional and off by default, allowing supported web assistants to access tools on the computer; before enabling it, you should review its data flow and provider policies. Permissions, previews, and confirmations vary by tool, and you should not assume a uniform confirmation dialog exists; review behavior and authorize before execution.
Regarding platform support, macOS 13+ and Windows 10+ (64-bit) are supported; Linux, iOS, and Android are not yet supported, and you can join the waitlist on the download page. For support and feedback, you can report issues through the feedback tool provided by LMCP or GitHub Issues (include the operating system and app version, and remove sensitive content from logs) or request features; the support email is ctpo@colibird.co, and for vulnerabilities see SECURITY.md. For release history and current installers, see GitHub Releases and the download page; Mac and Windows may have different release versions, and a GitHub tag itself is not a signal of a full deployment. For current availability and commercial terms, see the website and LICENSE; the README does not promise permanent pricing or future features. The core binary is proprietary software, and the documentation and scripts have their own licenses.
