This page states our project position; it is not a ranking of community projects. Other
desktop clients may better serve different user needs.See the desktop client comparison for the complete comparison.
An open-source organization without commercial bundling
We want this software to be maintained by an open, transparent, and sustainable open-source organization rather than becoming a bundle for commercial services, model channels, sponsorship promotion, or traffic acquisition. That means we aim to:- make product decisions for the software itself rather than conversion, promotion, or commercial partnerships;
- avoid presenting model gateways, paid services, or sponsorships as required product capabilities;
- state our relationship with DeepSeek, upstream Harness, and other community projects clearly;
- keep source, issue discussions, build processes, and release artifacts verifiable where possible;
- record any future change to the project boundary explicitly instead of silently changing the promise.
Quality before a community plugin bundle
Harness provides strong composition through “everything is a plugin,” but community plugins currently vary in maintenance, interaction design, compatibility, error handling, and security boundaries. Preinstalling many plugins can increase the feature count quickly, but it also distributes their quality and supply-chain risks to every user. We choose a restrained default:- bundle only the features and first-party integrations required for the desktop runtime that we can continuously validate and recover;
- keep optional plugins visibly optional instead of treating preinstalled quantity as software quality;
- prioritize installation, startup, updates, profiles, cores, logs, and failure recovery;
- explain plugin failures and offer recovery instead of exposing users directly to dependencies;
- consider uninstall, upgrade, rollback, cross-platform behavior, and long-term maintenance before adding a capability.
Stay light and pursue Codex-level responsiveness
Our goal is not only to open a window quickly. We want the desktop client to progressively reach the product standard demonstrated by Codex across feature completeness, interaction responsiveness, stability, detail consistency, and recovery. Tauri 2 is a means toward that goal, but not the only reason for the project:- reuse the system WebView instead of bundling another complete Chromium runtime;
- manage processes, downloads, updates, PATH, and OS integration in Rust;
- keep Harness as an independent local service while the desktop adds only necessary host capabilities;
- update the desktop and Harness core separately to avoid unnecessary replacement.
“Lightweight” is not only installer size. The Harness core, Node.js, pnpm, plugins,
startup time, memory, input latency, and recovery all affect the experience. We will
improve from measured results rather than claim performance leadership merely because we use Tauri.
Stay restrained and objective
DeepSeek Harness and its ecosystem are changing quickly. We cannot guarantee that every plan will ship, and we cannot promise upstream stability. We therefore want to promise only what we can verify and maintain within our current ability. Our principles include:- do not present planned features as current delivery commitments;
- do not substitute stars, plugin counts, or a feature checklist for reliability and user experience;
- do not state unmeasured “faster,” “safer,” or “lighter” claims as absolute facts;
- document unsupported platforms, known limitations, and failure modes clearly;
- describe other projects from source, releases, and test results without guessing their motives;
- reduce scope when our ability is insufficient and finish existing workflows first.