Back to catalog
ActivityWatch
MPL-2.0ActivityWatch is an open-source, local-first alternative to RescueTime that logs desktop and browser activity into an on-device SQLite database.
Open-source alternative to — compare it against:
Overview
ActivityWatch is an automated time tracker that records desktop application usage, active window titles, and browser tabs without sending telemetry off the host machine. It stores raw event streams in a local SQLite datastore queried through an embedded REST API. Background watchers collect user presence and application focus through periodic heartbeats rather than continuous screen capture or heavy background polling.
Why ActivityWatch instead of RescueTime
RescueTime runs a proprietary desktop daemon that aggregates application titles, categorizes productivity according to cloud heuristics, and uploads usage data to remote servers. That architecture introduces continuous network egress and cloud latency when querying historical timelines. ActivityWatch replaces that ingestion pipeline with on-device collection: watchers dispatch heartbeats over a local socket to aw-server, which writes directly into an embedded SQLite database. Because all timeline queries and categorization filters execute against local storage, dashboard rendering happens immediately without external API round-trips.
Workstation resource allocation and fleet aggregation define the trade-off. RescueTime offloads historical rollups to its cloud cluster, freeing local disk space but creating an ongoing per-seat subscription and potential data privacy risks for agencies handling sensitive client intellectual property. The compiled aw-server-rust backend keeps memory usage under one hundred megabytes, yet multi-workstation reporting remains decentralized. Each developer retains their timeline locally, making ActivityWatch an efficient personal diagnostic tool rather than a centralized employee surveillance platform.
What it does well
- Sub-second query execution: An embedded query engine processes raw bucket events directly on the local CPU, compiling daily and weekly timeline summaries without network latency.
- Low-overhead Rust server option: Running aw-server-rust caps resident memory usage at tens of megabytes, preventing background profiling from competing with compiler cycles.
- Event-driven heartbeat collection: Watchers record active window titles, browser URLs, and keyboard presence only on state changes, avoiding constant disk writes.
- Granular local data sovereignty: All tracked activity records stay inside an on-device SQLite database, keeping confidential codebases and client URLs off third-party clouds.
