ActivityWatch MCP Server

The ActivityWatch MCP Server bridges the gap between an AI assistant and your personal time‑tracking data. ActivityWatch is a free, open‑source tool that records how you spend your computer time—applications, windows, browser tabs, and more. By exposing ActivityWatch’s data through the Model Context Protocol (MCP), the server allows large‑language models such as Claude to ask natural‑language questions about your activity and receive structured answers without exposing raw logs or requiring manual exports.

At its core, the server offers four primary capabilities. First, it can list all buckets, giving a quick overview of the different data streams ActivityWatch collects (e.g., “window,” “web,” or “afk”). Second, it lets you run AQL queries—ActivityWatch’s powerful query language—so you can slice and dice your time‑tracking data for any period or metric. Third, the get raw events tool retrieves the underlying event records from a specified bucket, enabling deeper analysis or custom visualizations. Finally, get settings exposes the current ActivityWatch configuration, allowing an assistant to report or modify preferences on your behalf.

Developers and power users benefit from these features in several real‑world scenarios. An AI assistant can answer questions like “Which apps have I used the most today?” or “Show me my browsing history for this week,” drawing data directly from ActivityWatch without manual file handling. Teams can integrate the server into productivity dashboards, letting a virtual assistant generate weekly reports or alerts when time spent on non‑productive apps spikes. Because the server follows MCP standards, any client that supports MCP—Claude for Desktop, LangChain agents, or custom workflows—can seamlessly query the server using simple JSON commands.

What sets this MCP apart is its tight coupling with a proven time‑tracking platform and the ability to execute complex queries on demand. The server’s run-query tool supports caching via a query name, reducing redundant processing for repeated requests. The optional includeData flag in bucket listings lets developers decide whether they need just metadata or full data sets, optimizing bandwidth. Together, these features provide a flexible, secure, and efficient bridge between human intent expressed in natural language and the granular insights stored by ActivityWatch.