Maekon
Maekon is an Apache-2.0 local-first desktop agent. It can be used independently without ONESHIM, runs on your machine, and turns scattered local work signals into suggested next actions plus automation with approval rules without sending your raw activity to anyone else by default.
What it does
- Captures local work signals — active windows, screen frames (delta-encoded), system metrics, and (optional) input patterns. All processing happens locally.
- Detects work regimes — clusters your activity into focused sessions automatically; no manual time-tracking.
- Suggests candidates and gates actions — surfaces relevant next-action suggestions through local heuristics or explicitly configured streams, then routes approved actions through policy checks, sandbox boundaries, and audit logs.
- Integrates with your tools — connects to Linear, Jira, Slack, Notion, Calendar, and more (your tokens, your control).
When to use Maekon
- You want observable productivity data without giving up control of your activity stream.
- You're a developer / researcher / knowledge worker whose work signals live in many tools at once.
- You need on-device AI for security or compliance reasons (GDPR, ISMS-P, SOC 2 boundaries).
- You're comfortable running an open-source desktop client (macOS / Windows / Linux).
When not to use it
- You want a fully managed SaaS — Maekon is an independent local agent. Organization deployment through ONESHIM is a separate optional PoC path.
- You need a finished product — current release is
v0.0.1-rc.x(pre-release). APIs and data formats may shift.
Project at a glance
| Platforms | macOS (Universal), Windows (x64), Linux (x64) |
| Stack | Rust + Tauri 2 (desktop) · React 19 (UI) · SQLite local storage · optional gRPC/SSE connected path |
| License | Apache-2.0 |
| Source | github.com/pseudotop/maekon-client |
| Releases | latest releases |
| Landing | maekon.dev |
| Enterprise path | Optional ONESHIM edge input and control channel under explicit PoC scope |
Next steps
- Installation — Use the current verified prerelease install path for your platform.
- Getting Started — First run, local mode, and optional connected streams.
- Architecture — How the 15-package workspace fits together.