社区 Community 需审批 Needs approval 例行 Routine 多Agent Multi-agent Skill Skill

Personal Jarvis:开源本机语音编排桌面(Grok Bot 相邻) Personal Jarvis: open-source local voice-orchestrated desktop (Grok Bot-adjacent)

Personal Jarvis: open-source local voice-orchestrated desktop (Grok Bot-adjacent)

用本机语音桌面编排 Agent。 Orchestrate Agents through a local voice desktop.

Personal Jarvis:开源本机语音编排桌面(Grok Bot 相邻) 预览
无自托管预览图;以仓库 README 与官网为准。 No self-hosted preview image; refer to the repo README and the site.
打开原文Open source

详细介绍Details

Show HN https://github.com/PersonalJarvis/PersonalJarvis — Apache-2.0 本机桌面:Jarvis 用语音/文字编排 agents、coding session、桌面动作与 MCP;可挂例行与本地 Markdown wiki,产物进 Artifacts。作者在 HN 标题写成 Grok Bot 开源替代;README 强调跑在自己电脑上、模型自选(托管/本地)。收录为可复现的相邻栈,不当官方替代品。 supports routines and a local Markdown wiki, with outputs in Artifacts. The HN title frames it as an open-source Grok Bot alternative; the README stresses running on your own computer with your choice of hosted or local models. Catalogued as a reproducible adjacent stack, not an official replacement.

它是什么What it is

Personal Jarvis 是开源本机桌面应用,许可证 Apache-2.0(GitHub license API 于 2026-09-25 仍为 Apache-2.0)。README 写明:中心是 Jarvis 语音编排器,把对话接到 agents、coding workspace、桌面动作,以及你选的插件 / skill / MCP;也可以打字。工作区可建带说明与例行的 specialist,用 Jarvis Voice 向其他 App 听写,知识放本地 Markdown wiki,报告和文件进 Artifacts。模型由你选(托管、本地或混用)。Show HN item 49833292 的标题把它写成 Grok Bot 开源替代。本站按相邻栈收录:可自建、可核对,不是官方 Grok Bot。 Personal Jarvis is an open-source desktop app under Apache-2.0 (GitHub license API still reported Apache-2.0 on 2026-09-25). The README puts Jarvis, a voice orchestrator, at the center: conversation connects to agents, a coding workspace, desktop actions, and the plugins / skills / MCP you choose — typing works too. The workspace can hold specialists with their own instructions and routines, Jarvis Voice can dictate into other apps, knowledge can live in a local Markdown wiki, and reports and files land in Artifacts. You pick the models (hosted, local, or mixed). The title of Show HN item 49833292 frames it as an open-source Grok Bot alternative. This library lists it as an adjacent stack you can self-host and check — not official Grok Bot.

为什么推荐Why we recommend it 对照 README:本机桌面 + 语音编排 + agents/routines/MCP,不是「每席一云电脑」那条路;与 Errand / Open Bot 同属相邻对照,不当官方 Grok Bot 替代品宣传。 Per the README: local desktop + voice orchestration + agents/routines/MCP — not the per-seat cloud-computer path; listed beside Errand / Open Bot as an adjacent comparison, not marketed as an official Grok Bot replacement.

核验记录:来源、抓取与截图说明Verification log: sources, checks, and captures

Apache-2.0(2026-09-25 gh api,65★)。Show HN item 49833292。 (item 49833292,2026-09-24 ~16:43Z,Algolia 核对 1pt) 仓库于 2026-09-25 gh api:Apache-2.0、65★。 官网 personaljarvis.ai HTTP 200。Apache-2.0 (gh api, 2026-09-25, 65★). Show HN item 49833292. (2026-09-24 ~16:43Z, 1pt on Algolia at check time) Show HN item 49833292 at https://github.com/PersonalJarvis/PersonalJarvis — Apache-2.0 local desktop: Jarvis orchestrates agents, coding sessions, desktop actions, and MCP by voice or text; Repo on the 2026-09-25 gh api: Apache-2.0, 65★. Site personaljarvis.ai HTTP 200.

对照Contrast

左列是仓库 README 与官网对得上的事实。右列只说明为什么放在 Grok Bot Cases 旁边对照,不补写本库没核过的官方细节。 The left column is what the README and site actually say. The right column only says why it sits next to Grok Bot Cases — it does not fill in official details this library has not checked.

Personal Jarvis Personal Jarvis 对照 Grok Bot Beside Grok Bot

电脑在哪 Where the computer is

Personal Jarvis Personal Jarvis

跑在你自己的 Windows / macOS / Linux 桌面上;语音、桌面动作和本地文件都在本机。云模型只在你选用时才接到请求内容。 Runs on your own Windows / macOS / Linux desktop; voice, desktop actions, and local files stay on-device. Cloud models only receive request content when you choose them.

对照 Grok Bot Beside Grok Bot

Grok Bot 文档强调每席有自己的云电脑。这里是本机路线,对照的是「有桌面、能动手」,不是同一托管模型。 Grok Bot docs emphasize a per-seat cloud computer. This is the local path — compared on “has a desktop and can act,” not the same hosting model.

怎么交办 How you assign work

Personal Jarvis Personal Jarvis

语音或文字开聊;可建 specialist、挂例行、接 MCP / skill,并看 Artifacts 与 run history。 Start by voice or text; build specialists, attach routines, connect MCP / skills, and inspect Artifacts plus run history.

对照 Grok Bot Beside Grok Bot

用来对照「队友 + 可重复流程」。官方 Routine / Skill 名称以 xAI 文档为准。 Useful as a comparison for teammates plus repeatable flows. Official Routine / Skill names follow the xAI docs.

许可与边界 License and boundary

Personal Jarvis Personal Jarvis

Apache-2.0;托管模型与订阅另计费。README 写明本地并不等于所有集成离线。 Apache-2.0; hosted models and subscriptions bill separately. The README notes that local install does not make every integration offline.

对照 Grok Bot Beside Grok Bot

收录为相邻开源栈;不要当成已替你跑通的官方替代路径。 Catalogued as an adjacent open-source stack — not a verified official replacement path.

怎么试Try it

  1. 看仓库与许可 Check the repo and license

    打开 GitHub PersonalJarvis/PersonalJarvis,确认 LICENSE 仍是 Apache-2.0,并读 README 的隐私与本机数据说明。 Open GitHub PersonalJarvis/PersonalJarvis, confirm LICENSE is still Apache-2.0, and read the privacy / local-data notes in the README.

  2. 按官方安装脚本装桌面 Install the desktop with the official script

    README 给出 macOS/Linux 的 curl 安装脚本与 Windows 的 irm 脚本;装完后在应用里选语言、唤醒词和模型提供方。本站未代跑安装。 The README ships a curl installer for macOS/Linux and an irm script for Windows; finish language, wake phrase, and provider setup in the app. This site did not run the installer for you.

  3. 先试一条低风险对话 Try one low-risk conversation first

    先用文字 Chat 或短语音问一个不碰敏感账号的问题;确认审批与桌面控制开关符合你的预期,再接 MCP 或例行。 Start with text Chat or a short voice ask that avoids sensitive accounts; confirm approvals and desktop-control toggles match what you want before attaching MCP or routines.

边界Boundaries

  • 本站未替你安装或授权 Personal Jarvis;桌面控制与语音权限风险自担。 This site did not install or authorize Personal Jarvis for you; you own desktop-control and voice-permission risk.
  • HN 标题写「Grok Bot 替代」不等于功能对等;云电脑、审批闸与官方插件以 xAI 文档为准。 An HN title that says “Grok Bot alternative” is not feature parity; cloud computers, approval gates, and official plugins follow the xAI docs.
  • 托管模型会离开本机;本地模型能力受硬件与所选权重限制。 Hosted models leave the machine; local-model capability is limited by hardware and the weights you choose.

相关案例Related cases