Tag
作者分享了将桌面智能体的规划与执行分离的架构：由 Claude Code 等前沿模型批量规划步骤，小型快速模型 Jev 负责执行点击与操作，普通步骤跳过模型，从而显著提升速度与降低成本，测试中 12 个随机按钮达成 47/48、约 1.2 秒每次点击。代码已开源。
Garden Desk is a free, open-source desktop AI agent that runs entirely on your local machine to process local files, using no accounts, no telemetry, no cloud, and executing tools in an isolated no-network microVM with read-only source access.
Kent C. Dodds announces an upcoming video on improving personal software development processes, with a focus on adapting tools like AI models for daily problem-solving. John Lindquist shares examples of personal apps built for convenience and agentic features.
Alibaba Cloud integrates TinyFish's Mako execution engine into its desktop agent to deliver a web-native, workflow-trained intelligence layer for handling complex multi-step tasks with lower latency and fewer stalls.
Skales is a private, local AI desktop agent for Windows, macOS, Linux, and Android that performs autonomous tasks, supports multiple AI providers, and emphasizes privacy with offline capabilities.
A 6-week real-world experiment using an open-source desktop agent shell with a three-model split (Haiku triager, Sonnet reviewer, Opus executor) reports a 64% cost reduction and details failure modes like context bloat and runaway sub-agents.
Peter's article clarifies the fundamental boundary between cloud agents and desktop agents, discussing the security, runtime, and infrastructure challenges of cloud agents operating in unsupervised, shared hardware environments, and points out that agent runtime will increasingly resemble a small OS in the future.
Orkas is an open-source, local-first desktop agent app where a lead agent coordinates specialized sub-agents, each with its own context boundary, using user-provided API keys from various LLM providers.
Cui Tianyi (former quantitative specialist at Jane Street) joined DeepSeek in March 2025 as the head of the newly formed Harness team, developing a desktop agent product that competes with Anthropic's Claude Code.
A developer shares four common accessibility tree pitfalls that break desktop agents: stale PIDs after app switches, modal sheets intercepting clicks, multi-monitor coordinate issues, and silent failures. Solutions include detecting frontmost app changes, explicit modal checks, and correct coordinate targeting.
ByteDance open-sourced UI-TARS-desktop, a native desktop GUI agent with 31.4k GitHub stars that uses vision models to control local or remote applications via natural language. The tool runs locally for privacy, supports Windows and macOS, and includes a CLI with streaming output for developers.
andrewyng/aisuite is a lightweight Python library providing a unified Chat Completions API across multiple LLM providers and an Agents API with tools. It also includes OpenCoworker, a desktop AI agent for everyday tasks.