Release version 0.39.0

This one lets an agent write your commit messages, lets a command run on your own computer even from a remote folder, and lets codeg-server sign people in through a proxy such as Cloudflare Access instead of its access token.
Alongside that: Recent in the sidebar lists what you used last, on Windows an update waits for Restart to update and no longer stops at "Failed to kill codeg", the Linux AppImage no longer hands its own libraries to agents and the terminal, and a phone that comes back to the page gets the rest of the reply.

New

• An agent can write your commit message: the ✨ button by the commit box, in the Changes tab and the commit window, has the agent of the conversation you're in write one from exactly what will be committed, taking anything you've typed as a hint, and adds no conversation to your list. (#931 requested by @trueai-org, #928 by @Time-w)
• With untracked files around, the ✨ button first asks which of them the message covers, and Enter then commits exactly those, with a line under the commit box saying how many it will take.
• Commit message settings… in the More Git actions menu chooses who writes the message and how: a fixed agent, model and options, instructions it always follows such as the language or format, and committing as soon as it's written, for all folders or just one project.
• Manage commands has a Run on this computer switch, so a command such as opening the remote folder in your local VS Code or Zed over SSH runs on the computer with the desktop app, and Show in all folders keeps it at hand across folders, worktrees and remote connections. (#974, @fengpanhe)
• codeg-server can sign people in through a proxy such as Cloudflare Access: with CODEG_AUTH=proxy-jwt, a browser the proxy has let in no longer asks for the access token, which codeg-cli and scripts that bypass the proxy keep using, as the proxy sign-in guide explains. (#985, @MagTer)
• Grok's custom model card can declare the model's reasoning-effort levels, so the composer offers a reasoning-effort selector for a custom model too. (#993 requested by @HOMEFTW)

Improved

• Recent in the sidebar lists the conversations you used last instead of the ones created last, ten at a time, with each row showing when it was last active, and Chat rows show their last activity too. (#1010, @Flyneen)
• On Windows, a downloaded update waits for you to click Restart to update, as on macOS and Linux, instead of closing codeg the moment the download finishes, and an install that fails says so and can be tried again.
• The commit box in the Changes tab grows with the message, with Shift+Enter for a new line while Enter still commits.
• Clicking a delegation card unfolds the full task the sub-agent was given, and a resumed sub-agent's card shows there why it was resumed.
• The sidebar can be dragged narrower on Windows, Linux and in the browser, since only the macOS app needs room for its window buttons.
• Pi's model selector lists only the models your enabledModels setting allows. (#989, @xyalter)
• ZCode takes pasted and uploaded images and PDFs, and keeps each reply's token usage and time after you reopen a conversation. (#1003, #1005, @jackie-cqz)
• Claude Code moves to claude-agent-acp 0.94.0, and along with newer CodeBuddy, Cursor, Kimi Code, Qoder, ZCode and Command Code, a Claude Code fork sub-agent's card now shows its real report instead of "Fork started".

Fixed

• Updating codeg on Windows no longer stops at "Failed to kill codeg" while Windows or an antivirus is checking codeg.exe.
• In the Linux AppImage, the agents, terminals and other programs codeg starts get your system's own environment instead of the AppImage's, so on Ubuntu 26.04 installing and starting agents, npm and terminal commands no longer fail with "LIBSYSTEMD_254 not found". (#983 reported by @yanvigdev)
• A phone that puts the page in the background mid-reply gets the rest of the reply when it comes back, without reloading the page. (#386 reported by @wuzhaozhongguo)
• Behind a sign-in proxy such as Cloudflare Access, a page whose proxy sign-in has expired shows Sign-in expired with a Reload button, instead of trying to reconnect forever. (#984, @MagTer)
• A message sent on the canvas while the agent is still answering stays in the input with a notice, instead of disappearing. (#991 reported by @SeeSeaStar)
• A sub-agent waiting on you shows its permission, question or plan prompt in one card only, instead of one more card each time the agent checks on it.
• A new conversation no longer shows Load background activity right after its first message.
• An agent removed outside codeg, or lost with a switch to another Node version, no longer reads as installed in Settings. (#990 reported by @BillyOutlast)
• Grok installed with its own installer is found in ~/.grok/bin, also on Windows. (#994 reported by @HOMEFTW)
• The Codex settings no longer write options current Codex ignores, and the WebSocket switch is gone for Official Subscription, where it broke every turn, while Fast stays in the composer.
• The forge panel's open repository button works for a self-hosted GitLab or similar served over http, on its own port or under a subpath. (#997 reported by @run19910607)
• A pull-request task whose pull request was force-pushed reports only its own changes, on its card and in its done comment, instead of everything the base gained meanwhile.
• A project's .pi/mcp.json now asks for project trust before Pi starts the servers it lists.

Note

• Once 0.39.0 has run, 0.38.0 no longer starts on the same data, because this version adds to codeg's database: codeg-server's Roll back would restore a 0.38.0 that won't start, and so would reinstalling an older desktop app, so copy codeg's data folder before you update if you may want to go back.
• On Windows, the update from 0.38.0 still closes codeg to install as soon as the download finishes, because the version you update from decides that, and later updates wait for Restart to update.
• Run on this computer is only in the desktop app, and a remote folder's host has to be on 0.39.0 to save the switch on a folder's command: until then, turn on Show in all folders to keep the command on your computer.

Thanks to @MagTer, @fengpanhe, @Flyneen, @jackie-cqz and @xyalter for contributing to this release, and to @yanvigdev, @trueai-org, @Time-w, @wuzhaozhongguo, @SeeSeaStar, @BillyOutlast, @HOMEFTW and @run19910607 for the reports.

发布版本 0.39.0

这一版可以让智能体帮你写提交信息，远程文件夹里的命令也可以放到你自己的电脑上运行，codeg-server 还可以改用 Cloudflare Access 这类登录代理来登录，不必再输访问令牌。
同期还有：侧栏「最近」按最近使用排序；Windows 上的更新会等你点「重启以更新」再安装，也不会再卡在「Failed to kill codeg」；Linux AppImage 不再把自带的库带给智能体和终端；手机切回页面后能看到完整的回复。

新增

• 智能体可以帮你写提交信息了——「变更」标签页和提交窗口的提交框旁新增 ✨ 按钮，会让当前会话的智能体按这次实际要提交的内容写一条提交信息，已经输入的文字会作为提示，也不会在会话列表里多出会话。（#931 由 @trueai-org、#928 由 @Time-w 提出）
• 有未跟踪的文件时，✨ 按钮会先问你这条提交信息包括其中哪些，回车就只提交这些文件，提交框下方也会显示将带上几个未跟踪文件。
• 「更多 Git 操作」菜单里的「提交信息生成设置…」决定由谁、怎么写——可以指定智能体、模型和选项，加上每次都遵循的附加说明（比如用什么语言、什么格式），还能在写好后自动提交；可以对所有文件夹统一设置，也可以给某个项目单独设置。
• 「管理命令」新增「在本机运行」开关——比如用本机的 VS Code 或 Zed 通过 SSH 打开当前远程文件夹，这类命令会在运行桌面版的电脑上执行；打开「在所有文件夹中显示」后，切换文件夹、工作树或远程连接都能用到它。（#974，@fengpanhe）
• codeg-server 可以通过 Cloudflare Access 这类登录代理来登录——设置 CODEG_AUTH=proxy-jwt 后，已经通过代理登录的浏览器不用再输访问令牌，不经过代理的 codeg-cli 和脚本照常使用令牌，具体见代理登录说明。（#985，@MagTer）
• Grok 设置里的自定义模型可以声明支持的推理强度，输入框里也就有了这个模型的推理强度选择。（#993 由 @HOMEFTW 提出）

改进

• 侧栏「最近」按最近使用而不是创建时间排序，每次显示 10 个，每行显示的是最近活动时间；「对话」分组的每行也改为显示最近活动时间。（#1010，@Flyneen）
• 在 Windows 上，更新下载完后会等你点「重启以更新」再安装，和 macOS、Linux 一样，不会再一下载完就关掉 codeg；安装失败会提示，可以重试。
• 「变更」标签页的提交框会随内容变高，Shift+Enter 换行，回车仍然直接提交。
• 点开委派卡片能看到子智能体收到的完整任务，恢复执行的子智能体也会在这里显示恢复的原因。
• 在 Windows、Linux 和浏览器里，侧栏可以拖得更窄，只有 macOS 版需要给窗口按钮留出位置。
• Pi 的模型选择器只列出 enabledModels 设置允许的模型。（#989，@xyalter）
• ZCode 支持粘贴和上传的图片、PDF，重新打开会话后每条回复的用量和耗时也还在。（#1003、#1005，@jackie-cqz）
• Claude Code 升级到 claude-agent-acp 0.94.0，CodeBuddy、Cursor、Kimi Code、Qoder、ZCode 和 Command Code 也一并更新；Claude Code 的 fork 子智能体卡片现在显示它真正的结果，而不是「Fork started」。

修复

• 在 Windows 上更新 codeg 不会再停在「Failed to kill codeg」，即使这时 Windows 或杀毒软件正在检查 codeg.exe。
• Linux AppImage 版里，codeg 启动的智能体、终端和其他程序改用系统自己的环境，不再沿用 AppImage 的，在 Ubuntu 26.04 上安装和启动智能体、运行 npm 和终端命令不会再报「LIBSYSTEMD_254 not found」。（#983 由 @yanvigdev 反馈）
• 手机在回复途中把页面切到后台，切回来后能看到完整的回复，不用再刷新页面。（#386 由 @wuzhaozhongguo 反馈）
• 在 Cloudflare Access 这类登录代理后面，代理登录过期时页面会显示「登录已过期」和「刷新」按钮，不会再一直尝试重新连接。（#984，@MagTer）
• 在无限画布里给正在回答的对话发消息，消息会留在输入框里并给出提示，不会再消失。（#991 由 @SeeSeaStar 反馈）
• 子智能体等你处理时，权限请求、提问或计划审批只显示在一张卡片里，不会再每查询一次状态就多出一张。
• 新会话发出第一条消息后不会再马上出现「加载后台活动」提示。
• 在 codeg 之外卸载的智能体，或者切换 Node 版本后找不到的智能体，设置里不会再显示为已安装。（#990 由 @BillyOutlast 反馈）
• 用 Grok 官方安装程序装在 ~/.grok/bin 的 Grok 能被识别了，Windows 上也一样。（#994 由 @HOMEFTW 反馈）
• Codex 设置不会再写入新版 Codex 已不认的配置，官方订阅下也不再显示会让每轮对话都失败的 WebSocket 开关；快速模式仍在输入框里切换。
• 代码托管面板的「打开仓库」按钮支持走 http、用自定义端口或子路径的自建 GitLab 等服务。（#997 由 @run19910607 反馈）
• 处理拉取请求的看板任务在对方强推后，只统计自己的改动，卡片和完成评论里不会再算上基础分支期间新增的全部内容。
• 项目里的 .pi/mcp.json 现在要先信任项目，Pi 才会启动其中列出的服务。

注意

• 0.39.0 运行过之后，0.38.0 就无法在同一份数据上启动了，因为这一版给 codeg 的数据库加了内容：codeg-server 的「回滚」恢复出来的 0.38.0 起不来，重新安装旧版桌面版也一样；之后可能还要退回 0.38.0 的话，更新前请先复制一份 codeg 的数据目录。
• 在 Windows 上，从 0.38.0 更新的这一次仍会在下载完后直接关闭 codeg 安装，因为安装时机由更新前的版本决定，之后的更新会等你点「重启以更新」。
• 「在本机运行」只在桌面版里可用，远程文件夹的主机也要更新到 0.39.0 才能为文件夹命令保存这个开关；在此之前，打开「在所有文件夹中显示」就能把命令保存在你自己的电脑上。

感谢 @MagTer、@fengpanhe、@Flyneen、@jackie-cqz 和 @xyalter 为本次发布做出的贡献，也感谢 @yanvigdev、@trueai-org、@Time-w、@wuzhaozhongguo、@SeeSeaStar、@BillyOutlast、@HOMEFTW 和 @run19910607 的反馈。

Release version 0.38.0

This one adds codeg-cli, a command that runs agents through codeg from a terminal, a script or a CI job and prints the reply as text or JSON, while each run stays a conversation you can open and carry on in codeg.
Alongside that: a browser picks up the server's new release after an upgrade, /deny in a chat channel no longer gives Claude Code a lasting permission, tables and code blocks have one border again with code highlighted in light mode, codeg restarts into the Dock icon you pinned on macOS, and installing or updating on Windows no longer hangs while stopping codeg-mcp.

New

• codeg-cli runs agents through codeg from a terminal, a script or a CI job: codeg-cli -a claude_code -p "Review the latest commit" hands the prompt to the desktop app or codeg-server already running on this computer, or to a codeg-server elsewhere with --url and its token, and prints the reply as text, JSON or a stream of JSON events with documented exit codes, while the run is a conversation in codeg like any other. (#694 requested by @yasin1151)
• --permissions decides how a run with no one watching answers the agent's permission requests: refuse (the default), allow reads only, allow, end the run, or leave them for you in codeg, and it only ever answers for this once, never with an "always" that would write a lasting rule into the agent's settings.
• A run can go on in the background and be collected later with --detach, result and cancel, work in a git worktree of its own with --worktree, or go to several agents or assistants at once, and --output-schema has the agent hand back JSON that matches a schema you give.
• codeg-cli also lists and installs agents, lists your assistants and runs saved automations, and --standalone gives a single command, in a CI job for example, a codeg-server of its own that goes away when the command ends.
• Settings → System → Command-line access is the switch for codeg-cli on this computer and shows a command to try, while in the desktop app Install command-line tool puts codeg-cli on your PATH, and the app's own terminals and agents find its codeg-cli first.
• Conversations started from the command line carry a terminal mark in the sidebar, and Show in Recent gains Command line to list only them.
• codeg-cli comes with the desktop app, the server packages and the Docker image, install.sh and install.ps1 put it next to codeg-server, and from now on codeg-server's in-place update keeps it up to date. The command-line guide covers every command and option.

Improved

• Local images also show in permission requests, plan approvals and a to-do's Result, as they do in replies.

Fixed

• A browser picks up the server's new release after an upgrade, instead of sometimes opening the previous one from its cache, and a tab left open through the upgrade offers to reload. (#980 reported by @wuzhaozhongguo)
• /approve and /deny in a chat channel answer just this once and pick the answer by what it does, not by where it sits in the list: for a call Claude Code flags as risky, /deny could give it a lasting permission while /approve and automatic approval refused it. A request with no one-time answer now waits for you in codeg.
• Tables and code blocks in replies, thinking and Markdown previews have a single border again, with a code block's language, download and copy buttons in a bar above the code, and code in light mode is highlighted again.
• A table opened full screen keeps its exit button on screen.
• On macOS, codeg comes back after an update or a restart in the Dock icon you pinned, instead of as a second icon next to it.
• Installing or updating codeg on Windows no longer hangs at "Stopping codeg-mcp and codeg-computer-helper processes…" on a PC where Windows Management Instrumentation (WMI) doesn't answer, and install.ps1 waits at most 15 seconds for it.
• Two sub-agents delegated at the same moment into a folder codeg hadn't seen yet both start, instead of one of them failing.

Note

• Once 0.38.0 has run, 0.37.0 no longer starts on the same data, because this version adds to codeg's database: codeg-server's Roll back would restore a 0.37.0 that won't start, and so would reinstalling an older desktop app, so copy codeg's data folder before you update if you may want to go back.
• codeg-server updated in place from 0.37.0 doesn't get codeg-cli, since 0.37.0's updater knows nothing of it: take codeg-cli from the server package on this page and put it next to codeg-server. The desktop app's update, install.sh, install.ps1 and the Docker image all bring it.
• Command-line access is on by default: any program running under your account can use codeg-cli to run agents through codeg, as it could already read codeg's data, so switch it off under Settings → System → Command-line access if you don't use codeg-cli.
• On macOS, the update to 0.38.0 may still leave a second codeg icon in the Dock if codeg is pinned there, because the version you update from does that restart: quit codeg and open it from the pinned icon, and later updates aren't affected.
• On Windows, codeg-cli, like codeg-server, needs the Microsoft Visual C++ Redistributable to start.
• On Linux, codeg-cli, like codeg-server, needs glibc 2.34 or later and OpenSSL 3, which Ubuntu 22.04, Debian 12, RHEL 9 and later have, and there's no build for Alpine.

Thanks to @yasin1151 and @wuzhaozhongguo for the reports.

发布版本 0.38.0

这一版新增了命令行工具 codeg-cli：可以在终端、脚本或 CI 任务里通过 codeg 运行智能体，回复以文本或 JSON 输出，每次运行都是一个可以在 codeg 里打开、接着聊的会话。
同期还有：服务器升级后浏览器会用上新版本，消息渠道里的 /deny 不会再给 Claude Code 永久放行，表格和代码块恢复成一层边框、浅色模式下的代码重新有了语法高亮，macOS 上重启后会回到你钉在程序坞上的图标，Windows 上安装或更新不会再卡在停止 codeg-mcp 那一步。

新增

• 新增 codeg-cli，可以在终端、脚本或 CI 任务里通过 codeg 运行智能体——codeg-cli -a claude_code -p "Review the latest commit" 会把提示词交给本机正在运行的桌面版或 codeg-server，加上 --url 和令牌也能交给别处的 codeg-server；回复以文本、JSON 或 JSON 事件流输出，退出码都有文档说明，每次运行都是 codeg 里的一个普通会话。（#694 由 @yasin1151 提出）
• 用 --permissions 决定无人值守的运行怎么回应智能体的权限请求——拒绝（默认）、只允许读取、允许、结束运行，或者留给你在 codeg 里处理；每次都只回应这一次，从不选会在智能体设置里写下永久规则的「始终」。
• 运行可以转到后台、稍后再取结果（--detach、result、cancel），也可以用 --worktree 在单独的 git worktree 里进行，或者一次发给多个智能体或助手；--output-schema 能让智能体按你给的 JSON Schema 交回结果。
• codeg-cli 还能列出和安装智能体、列出助手、运行已保存的自动化；--standalone 会为单条命令（比如 CI 任务里的一条）启动一个专用的 codeg-server，命令结束就随之关闭。
• 「设置 → 系统 → 命令行访问」是本机 codeg-cli 的开关，还会给出一条可以直接试的命令；桌面版在这里另有「安装命令行工具」按钮，把 codeg-cli 放到 PATH 上，桌面版自带的终端和智能体也会优先用它自己的 codeg-cli。
• 从命令行启动的会话在侧栏带有终端标记，「最近显示」新增了「命令行」，可以只列出这些会话。
• 桌面版、服务器安装包和 Docker 镜像都自带 codeg-cli，install.sh 和 install.ps1 会把它装在 codeg-server 旁边，以后 codeg-server 的在线升级也会一并更新它；所有命令和选项见命令行文档。

改进

• 权限请求、计划审批和待办任务「结果」里的本地图片也能显示了，和回复里的一样。

修复

• 服务器升级后浏览器会用上新版本，不会再有时从缓存里打开旧版本，升级时一直开着的页面也会提示你刷新。（#980 由 @wuzhaozhongguo 反馈）
• 消息渠道里的 /approve 和 /deny 只回应这一次，并按选项的含义而不是在列表里的位置来选——以前遇到 Claude Code 标为有风险的调用，/deny 可能反而永久放行，/approve 和自动批准却会拒绝。没有「仅这一次」选项的请求，现在会留给你在 codeg 里处理。
• 回复、思考和 Markdown 预览里的表格和代码块恢复成一层边框，代码块的语言、下载和复制按钮放在代码上方的标题栏里，浅色模式下的代码也重新有了语法高亮。
• 全屏查看表格时，退出按钮不会再跑到屏幕外。
• 在 macOS 上，更新或重启后 codeg 会回到你钉在程序坞上的那个图标，不会再在旁边多出一个。
• 在 Windows 上安装或更新 codeg 不会再卡在「Stopping codeg-mcp and codeg-computer-helper processes…」——以前在 WMI 服务无响应的电脑上会一直等下去；install.ps1 查询 WMI 也最多等 15 秒。
• 智能体同时把两个子任务委派到 codeg 还没记录过的同一个文件夹时，两个都能正常启动，不会再有一个失败。

注意

• 0.38.0 运行过之后，0.37.0 就无法在同一份数据上启动了，因为这一版给 codeg 的数据库加了内容：codeg-server 的「回滚」恢复出来的 0.37.0 起不来，重新安装旧版桌面版也一样；之后可能还要退回 0.37.0 的话，更新前请先复制一份 codeg 的数据目录。
• 从 0.37.0 在线升级的 codeg-server 拿不到 codeg-cli，因为 0.37.0 的升级程序还不认识它：请从本页的服务器安装包里取出 codeg-cli，放到 codeg-server 旁边。桌面版的更新、install.sh、install.ps1 和 Docker 镜像都会带上它。
• 命令行访问默认开启——以你的账户运行的任何程序都能借 codeg-cli 通过 codeg 运行智能体，就像它们本来就能读取 codeg 的数据一样；用不到 codeg-cli 的话，可以在「设置 → 系统 → 命令行访问」里关掉。
• 在 macOS 上，如果把 codeg 钉在了程序坞，更新到 0.38.0 的这一次仍可能多出一个 codeg 图标，因为这次重启由更新前的版本执行：退出 codeg 后从钉住的图标重新打开即可，之后的更新不受影响。
• 在 Windows 上，codeg-cli 和 codeg-server 一样，要装了 Microsoft Visual C++ 可再发行程序包才能启动。
• 在 Linux 上，codeg-cli 和 codeg-server 一样需要 glibc 2.34 及以上和 OpenSSL 3，Ubuntu 22.04、Debian 12、RHEL 9 及更新的系统都满足，暂时没有适用于 Alpine 的版本。

感谢 @yasin1151 和 @wuzhaozhongguo 的反馈。

Release version 0.37.0

This one adds TRAE CLI and Command Code as built-in agents, lets a Claude Code session continue on your phone with Remote Control, and has Claude Code suggest your next prompt. The side panel gains Automations, To-dos, Issues and Pull requests tabs and now holds Agent Plan and Message navigation, the model and option pickers under the message box merge into one button, and Claude Code moves up to 0.93.0 along with seven more bundled agents.
Alongside that: a sub-agent waiting on you can be answered right where it was delegated, images in replies show even when they are outside the conversation's folder, a reply that arrives just as a turn ends is no longer lost, switching Cline's model mid-conversation takes effect, saving a Cursor MCP server in codeg keeps Cursor's own settings, and computer use moves to cua-driver 0.36.0.

New

• ByteDance's TRAE CLI is now a built-in agent: codeg runs the traecli you installed, with its mode, model and reasoning effort pickers in the message box, its MCP servers and skills managed from codeg's settings, delegation, and history for the conversations you run in codeg, while its settings page offers Upgrade and an optional personal access token for servers and Docker. (#933 requested by @lwxk)
• Command Code is now a built-in agent: install it from Settings → Agents and sign in once with command-code login or save an API key, and codeg runs it with its mode, model and effort pickers in the message box, its MCP servers and skills managed from codeg's settings, delegation, and history for the conversations you run in codeg. (#951 requested by @terranc)
• Claude Code's Remote Control works in codeg: send /remote-control in a Claude Code conversation, and a strip above it shows the session's claude.ai/code link and a QR code for picking it up in the Claude app or a browser, with a Turn off button, while the session stays connected even after you close its tab.
• Claude Code suggests your next prompt: after a turn, the empty message box shows in grey what Claude expects you to ask next, and Tab fills it in without sending.
• The side panel gains Automations, To-dos, Issues and Pull requests tabs for the project of the conversation you're in, where you can check its automations and their runs, work through its to-do board, and browse its issues and pull requests (merge requests on GitLab), with Open full page one click away.
• Agent Plan and Message navigation move from the corner of the conversation into the side panel's Session Details tab, which opens with a card showing the title, agent, status, model, Git branch and copyable session IDs, while the plan shows its progress step by step, even when folded.
• Message navigation follows the message you're reading, and clicking a message jumps there and opens it to show the sub-agents its round ran, as cards that open their conversations, and the files it changed, each with an inline or side-by-side diff.
• A sub-agent that needs you can be answered in the conversation that delegated to it: its "Waiting for task result" row opens by itself with the permission, question or plan approval inside, and its badge turns from awaiting approval to checked once you answer. (#973 reported by @lwxk)
• The model, mode and other options under the message box merge into one button that lists the current settings separated by "/", leaving out options still on their default, and clicking it opens one panel where you change each setting in place, with search for long lists such as models.
• The automation and to-do editors show an agent's settings with the same button and panel, instead of a dropdown for each option.
• A folder group in the sidebar can open or clone a folder straight into itself: hover its heading for Clone Repository and Open Folder, and the new folder lands in that group, which expands to show it.

Improved

• Updated bundled agents: Claude Code ACP 0.93.0 (Claude Code 2.1.300), OpenCode 1.19.3, Cline 3.0.73, CodeBuddy 2.168.2, Grok 1.0.62, Cursor 2026.10.16, Qoder 1.1.73, Antigravity 1.4.1.
• Computer use moves to cua-driver 0.36.0, which codeg downloads the next time it is needed, and on Linux taking a screenshot no longer forces an agent to read the window again before it can click.
• The sidebar's tag filter moves into View options as its first entry, in place of the funnel button at the top of the sidebar.
• Clicking the side panel tab that is already showing hides the panel, and on phones tapping the selected tab closes the drawer.
• A new automation starts in the folder the Automations page is filtered to, instead of always the first open folder.
• The sidebar no longer marks conversations whose worktree was removed, and hovering one still shows its original path.
• The built-in model catalog is refreshed with 8,897 models from 230 providers, up from 8,894.

Fixed

• Images in a reply show even when they are outside the conversation's folder, such as a login QR code an agent saved in /tmp, instead of a broken-image icon. (#978 reported by @wuzhaozhongguo)
• In the desktop app, a sub-agent that is already waiting on you shows its request in a window that reloads or opens later, instead of the request never reaching it. (#973 reported by @lwxk)
• A reply that arrives just as a turn ends stays in that turn, instead of occasionally going missing with the turn reported as ending "without producing any response".
• Expanding a to-do's progress or completion report in its conversation shows the full text, formatted, with a line saying whether it was recorded on the to-do, instead of a bare {"recorded":true} from Claude Code and Codex.
• Switching Cline's model after the first message now takes effect, as does the model codeg reapplies when you reopen a Cline conversation, instead of requests staying on the old model while the picker showed the new one.
• Cline stays on the version codeg installed, instead of updating itself to its latest release once a conversation closes.
• Cline conversations keep the time they actually ended, instead of jumping to the top of the sidebar with the time Cline was next started.
• Saving a Cursor MCP server in Settings → MCP, or installing it again from the MCP Marketplace, keeps the settings Cursor adds itself, such as envFile, auth and tls, and an SSE server stays SSE instead of turning into HTTP.
• Cursor-only MCP settings no longer spread to other agents when codeg writes a server into their configs, where Cursor's auth made OpenClaw reject its whole config and pi skip the server.
• Slash commands with a colon in their name now show as one command in the message box, such as Qoder's new sites: commands, and reach the agent exactly as named.
• On the Automations page, the detail pane follows the folder filter instead of showing an automation from another folder, and a folder closed in the sidebar shows its name in the filter instead of a number.

Note

• TRAE CLI 2.0 is available only on TRAE Enterprise's flagship plan, and codeg runs the traecli you install with TRAE's own installer, signed in with traecli login.
• Command Code needs Node.js 22 or later, and models you add to Command Code yourself can't be picked in the message box, as it lists only its own catalog there.
• TRAE CLI's and Command Code's first integrations leave some things out: conversations from their own terminals can't be imported yet, one you set up as a custom agent stays separate with its own history, and so far both have been tested on macOS only.
• With Claude Code ACP 0.92.0 and later, Claude Code makes one more request after most turns to predict your next prompt, mostly read from the prompt cache but still counted in your usage, while automations and chat channel bots skip it.
• Remote Control needs Claude Code signed in with a claude.ai subscription and talking to api.anthropic.com, so it isn't available through a gateway or a custom ANTHROPIC_BASE_URL.
• The Session Details tab no longer shows token usage or the created and updated times, which the Session Details dialog, opened from a conversation's menu, still shows.
• Cline applies Plan or Act only when it starts working on a conversation, so switching modes after the first message has no effect until Cline changes that.
• CodeBuddy 2.168.2 runs every step of a conversation started in its own terminal through a single REPL tool, so such a conversation shows one REPL card per step in codeg, while conversations you start in codeg aren't affected.
• On macOS, a codeg that opens at login may not come back after an update started from 0.35.2 or earlier, so click its Dock icon to open it, while updates started from 0.36.0 or later aren't affected.

Thanks to @wuzhaozhongguo, @lwxk and @terranc for the reports.

发布版本 0.37.0

这一版新增了 TRAE CLI 和 Command Code 两个内置智能体，可以用 Claude Code 的远程控制在手机上接着聊同一个会话，Claude Code 还会预测你的下一条提示词。右侧面板新增了自动化、待办任务、议题和拉取请求四个标签页，计划任务和消息导航也挪进了面板的「会话详情」，输入框下方的模型和各项选项合并成了一个按钮，Claude Code 升级到 0.93.0，另有七个内置智能体更新了版本。
同期还有：等你处理的子智能体可以直接在发起委派的会话里批准，回复里会话文件夹以外的图片也能显示了，回合结束瞬间到达的回复不会再丢失，Cline 会话中途切换模型能生效了，在 codeg 里保存 Cursor 的 MCP 服务器会保留 Cursor 自己的设置，电脑操作改用 cua-driver 0.36.0。

新增

• 字节跳动的 TRAE CLI 成为内置智能体——codeg 直接运行你安装好的 traecli，输入框里可以选模式、模型和推理强度，MCP 服务器和技能可以在 codeg 的设置里管理，支持委派，在 codeg 里进行的会话也会留下历史记录；设置页里可以「升级」，还能填写可选的个人访问令牌，方便在服务器和 Docker 上使用。（#933 由 @lwxk 提出）
• Command Code 成为内置智能体——在「设置 → 智能体」里安装，用 command-code login 登录一次或填上 API 密钥后，codeg 就能运行它：输入框里可以选模式、模型和推理强度，MCP 服务器和技能可以在 codeg 的设置里管理，支持委派，在 codeg 里进行的会话也会留下历史记录。（#951 由 @terranc 提出）
• Claude Code 的远程控制（Remote Control）可以在 codeg 里用了——在 Claude Code 会话里发送 /remote-control，会话上方会出现一条提示栏，显示本会话的 claude.ai/code 链接和二维码，可以在 Claude App 或浏览器里接着聊，栏上还有「关闭」按钮；远程控制开着时，即使关掉标签页，会话也会保持连接。
• Claude Code 会预测你的下一条提示词——回合结束后，空着的输入框里会用灰字显示 Claude 猜你接下来要说的话，按 ...

Release version 0.36.0

This one adds Kiro CLI and ZCode as built-in agents and introduces Assistants, ready-made delegation setups with their own agent, options and standing instructions that you @-mention to hand off a task. PlantUML code blocks now draw as diagrams, a to-do that needs you can be set aside with Handle later, the side panel's tabs move to a vertical rail you can reorder, and Claude Code moves up to 0.91.0 along with seven more bundled agents.
Alongside that: an expanded Read card shows the image it read, replies in a tab left open show their full working time, stopping a Claude Code turn no longer leaves its tool cards reading Completed, Antigravity's / menu lists your skills, Windows in-app updates no longer stop on codeg-mcp.exe, delegation keeps working when codeg is started from inside an agent session, and updated components close several security holes.

New

• Amazon's Kiro CLI is now a built-in agent: codeg runs the kiro-cli you installed with Kiro's own installer, with its agent, model, thinking and effort pickers in the message box, its skills and prompts in the / menu, history import, and a settings page with Upgrade, a permission mode and an optional API key. (#851, @keh4l)
• Z.ai's ZCode is now a built-in agent: install its ACP bridge in Settings → Agents and codeg runs your signed-in official ZCode through it, with its mode, model and thinking pickers in the message box, permission requests, stdio MCP servers, delegation, and history for the conversations you run in codeg. (#941, @jackie-cqz; #685 requested by @nn200433)
• Assistants are ready-made delegation setups you can @-mention: under Settings → Collaboration → Assistants, give one a name, avatar and description, pick an agent with every one of its options and add standing instructions, and mentioning it in the message box hands it the task with exactly that setup.
• Agents can also pick an assistant on their own when they delegate, going by its description, while delegation cards, Session Details and the sidebar show which assistant ran a task, and Configuration sync carries your assistants to your other machines.
• PlantUML diagrams are drawn from code blocks marked plantuml or puml in replies, thinking and file previews, in the same card as Mermaid diagrams with zoom, fullscreen, copy and download as SVG, PNG or .puml, all rendered on your machine. (#968 requested by @f1603849451-arch)
• A to-do that needs you can be set aside with Handle later: a task waiting for review or a failed one moves to Done as it is and leaves the badge count, and from there Follow up sends it back to the agent in the same session, a failed one can still be retried, and Back to Needs you returns it.
• The side panel's tabs move to a vertical icon rail along its right edge, where you can drag them into the order you like or move a focused one with Alt+↑/↓, and every window follows the new order.

Improved

• Updated bundled agents: Claude Code ACP 0.91.0 (Claude Code 2.1.300), OpenCode 1.19.1, Cline 3.0.72, Hermes 0.22.2, CodeBuddy 2.168.0, Grok 1.0.60, Cursor 2026.10.14, Qoder 1.1.72.
• Agent defaults under Settings → Collaboration now works like the assistant editor: one agent picker marks agents that aren't installed and counts your overrides, options left on Default are dimmed, you can change the model while its other options load, and a saved value the agent no longer offers shows as unavailable instead of blank.
• Updates are checked against the version they were signed for, in the desktop app and in the server's online upgrade alike, so an older signed package can't pass for a newer one.
• The built-in model catalog is refreshed with 8,894 models from 230 providers, up from 8,831.

Fixed

• An expanded Read card for an image now shows the picture, as when Codex views one, also for images outside your workspace folders such as in /tmp, and you can open it full size, copy or download it, while a reopened Codex conversation shows the same card instead of a block of base64 text. (#977 reported by @wuzhaozhongguo)
• Antigravity's / menu lists your skills after its own commands, from the same folders Settings → Skills manages, and so do the automation and to-do editors. (#971 reported by @lwxk)
• Several security issues in bundled components are fixed: pasting HTML into the message box can no longer run a script, a Mermaid diagram can no longer place content outside its own picture, and a math formula can no longer become a live javascript: link.
• Windows in-app updates no longer stop on "Error opening file for writing" for codeg-mcp.exe, as codeg now shuts its agents down before the installer starts, and the installer moves aside a file that is still in use.
• Delegation keeps working when codeg is started from inside one of its own agent sessions, as when an agent relaunches it, instead of every delegation failing with "No such file or directory". (#966, @Liang-HZ)
• A codeg started that way also keeps its own temporary folder instead of deleting it at startup, which had left its terminals and Pi's model list without one.
• In a conversation tab you keep open across several prompts, each reply shows its full working time and token usage, instead of a reply that ran for an hour and a half reading 11 s until you reopened the conversation.
• After you stop a Claude Code turn, the tool cards it left open show how they ended, sub-agents' calls included, instead of reading Completed until you reload.
• A resumed Claude Code conversation no longer shows its earlier history a second time, as a copy without durations that doubled the message count and pulled tool results onto the wrong calls.
• A Claude Code turn that starts on its own after a background task finishes is timed from that moment, instead of counting the idle time before it.
• A Claude Code background sub-agent that finishes while its own background work still runs no longer reads as done: its card labels the result interim, and no "finished" notification goes out until the real end.
• A failed Claude Code tool call no longer shows stray code-fence marks around its error while live, so it reads as it does after a reload.
• Claude Code's background task results show <, > and & as written, instead of <, > and &.
• CodeBuddy conversations reopened from history show a slash command as you typed it, instead of its tags and the prompt it expands to.
• Inline math in a reply no longer stays as raw $$ text when the reply is drawn before the math renderer has loaded.
• Moving conversation tags twice in quick succession no longer loses a move, and a window that opens just as a tag changes elsewhere picks up the change.
• The remote workspace menus, connections dialog and connection error page are now translated into Traditional Chinese, Japanese, Korean, Spanish, German, Portuguese and Arabic, instead of showing mostly in English.
• Searching remote workspace connections now says "No matching remote connections" when nothing matches, instead of repeating the search box's prompt.
• In Arabic, tabbed pages now lay out right to left like the rest of the window.

Note

• Kiro and ZCode run the CLIs you install yourself: Kiro needs kiro-cli from Kiro's official installer, and ZCode needs official ZCode installed and signed in plus Node.js 22 or later with node:sqlite, since codeg installs only ZCode's ACP bridge.
• A Kiro CLI you added from the ACP registry as a custom agent becomes the built-in Kiro on upgrade, with its conversations, tabs, settings, automations, to-do settings, delegation defaults, tags and message box picks carried over.
• ZCode Start Plan accounts can't run through the bridge, as their requests need a captcha only the ZCode desktop app can solve, while GLM Coding Plan accounts are not affected.
• ZCode's first integration leaves some things out: forking from a message, attachments, HTTP MCP servers and importing ZCode's own history aren't supported yet, a ZCode you set up as a custom agent stays separate with its own history, and so far it has been tested on Windows only.
• Assistants work only with delegation on, under Settings → Collaboration → General → Enable delegation, which is off by default.
• Cline doesn't get codeg's collaboration tools, such as delegation, as it never connects MCP servers handed to it over ACP, while the MCP servers in its own settings keep working.
• OpenCode 1.19.1 discontinued session sharing, so the Sharing setting on its settings page no longer has any effect.
• PlantUML diagrams render without a PlantUML server, so standard-library includes such as <C4/...> aren't available.

Thanks to @keh4l, @jackie-cqz and @Liang-HZ for contributing to this release, and to @wuzhaozhongguo, @lwxk, @f1603849451-arch and @nn200433 for the reports.

发布版本 0.36.0

这一版新增了 Kiro CLI 和 ZCode 两个内置智能体，还带来了「助手」：把智能体、选项和常驻指令存成一套委派预设，在输入框里 @ 它就能把任务交出去。PlantUML 代码块会画成图了，等你处理的待办任务可以先「以后处理」，右侧面板的标签页移到了可以拖动排序的竖排图标栏，Claude Code 升级到 0.91.0，另有七个内置智能体更新了版本。
同期还有：展开的 Read 卡片会显示读到的图片，一直开着的标签页里每条回复都会显示完整的工作时长，停止 Claude Code 的回合后工具卡片不会再显示「已完成」，Antigravity 的斜杠菜单会列出你的技能，Windows 应用内更新不会再卡在 codeg-mcp.exe 上，在智能体会话里启动 codeg 后委派也能正常使用，更新后的依赖组件还修复了几处安全问题。

新增

• 亚马逊的 Kiro CLI 成为内置智能体——codeg 直接运行你用 Kiro 官方安装程序装好的 kiro-cli，输入框里可以选 Kiro 的 agent、模型、思考开关和思考强度，斜杠菜单里有它的技能和提示词，支持导入历史会话，设置页里可以「升级」、选择权限模式，还能填写可选的 API Key。（#851，@keh4l）
• 智谱的 ZCode 成为内置智能体——在「设置 → 智能体」里安装它的 ACP 桥接器后，codeg 会通过它驱动你已登录的官方 ZCode，输入框里可以选模式、模型和思考选项，支持权限请求、stdio MCP 服务器和委派，在 codeg 里进行的会话也会留下历史记录。（#941，@jackie-cqz；#685 由 @nn200433 提出）
• 新增「助手」，可以 @ 的委派预设——在「设置 → 协作 → 助手配置」里给助手起名、选头像、写描述，选好智能体并定下它的每一项选项，再写上常驻指令；在输入框里 @ 它，任务就会按这套设置交给它处理。
• 智能体自己委派时也能按描述挑选助手，委派卡片、会话详情和侧栏会显示任务是由哪个助手完成的，「配置同步」也会把助手带到你的其他设备上。
• PlantUML 图表可以直接画出来了——回复、思考过程和文件预览里标为 plantuml 或 puml ...

Release version 0.35.2

This one opens PDF files right in a file tab, lets the repository panel turn issues and pull requests into to-dos on its own, and flashes the taskbar or bounces the Dock when an agent needs you. Token Usage now estimates what your tokens would cost at API list prices, and Claude Code moves up to 0.90.0 and Kimi Code to 2.2.0, along with six more bundled agents.
Alongside that: a draft can switch agents while it is still connecting, built-in browser pages on Linux stay inside the workspace window, Office previews keep their place across tab switches and wait for slow documents, message navigation reaches the whole conversation, the back gesture on phones closes what you opened instead of leaving codeg, and agents stop polling for your feedback in a loop.

New

• PDF files open in a built-in viewer in file tabs, the conversation's file viewer and canvas file cards, with zoom, fit to width, a page field, find, and a password prompt for protected files.
• The repository panel can turn issues and pull requests into to-dos on its own: under Panel settings → Auto-processing, have it scan on a schedule or on webhook events from GitHub, GitLab, Gitea or Forgejo and pick items by label and author, and it handles each item only once.
• The desktop app flashes its taskbar button or bounces its Dock icon when an agent needs you while you are in another window, on the same events as desktop notifications such as a finished turn, a permission request or a question, and Settings → General → Taskbar and Dock alerts sets each event to Off, Briefly or Until you return.
• Token Usage estimates what your tokens would cost at API list prices, for the selected range and for each model, with input, output, cache reads and cache writes priced separately from the models.dev catalog. (#503, @Adam-Dalloul)
• You can choose which tabs the @ mention panel shows and in what order under Settings → General → Mention panel, in the message box and the to-do and automation editors alike, with the Files tab always shown.
• To-dos can be searched by keyword, and #12 finds the task or issue with that number.
• The file preview header has a download button next to its ⋮ menu in the web version and in remote workspace windows, and it works for files outside your workspace folders too, such as an agent's output in /tmp. (#953, @ker2xu)

Improved

• Updated bundled agents: Claude Code ACP 0.90.0 (Claude Code 2.1.296), Kimi Code 2.2.0, OpenCode 1.19.0, Antigravity 1.4.0, Hermes 0.22.0, Cline 3.0.71, Grok 1.0.52, Qoder 1.1.71.
• Issue tasks check the report before acting: with Fix / implement or Plan first, when the report can't be reproduced, is already fixed, comes from the reporter's environment or describes intended behavior, the agent changes no code, writes no plan and gives its verdict in the first line of its reply.
• Message navigation covers the whole conversation, listing every message with its number and the files it changed, and picking one that hasn't loaded yet loads history up to it and jumps there.
• A draft conversation can switch to another agent while the one you picked is still connecting, in chat tabs and canvas draft cards alike, and the last agent you pick is the one that connects, with any startup error shown in the message box. (#963, @ker2xu)
• Sub-agent cards in Claude Code and Qoder show the sub-agent's thinking, text and tool calls in the order they happened, with thinking folded as in the main conversation, and a finished sub-agent folds all of it into one "Sub-agent activity" line with its duration. (#917, @Liang-HZ)
• When a Qoder turn breaks off while its sub-agent runs to the end, the sub-agent's card shows the result it finished with. (#917, @Liang-HZ)
• Qoder can ask you questions with its own question tool, on the same question card other agents use. (#915, @Liang-HZ)
• The model list under Agent defaults in Settings → Collaboration is grouped by provider, like the model picker in the message box, so same-named models from two providers can be told apart. (#947, @Liang-HZ)
• The web version caches itself in your browser, so a tab the browser put to sleep reloads without downloading codeg again, and a reload while the server is restarting or offline shows codeg's connection screen instead of the browser's error page. (#924, @leehom0123)
• Claude Code's Read card shows an "Over size limit" badge when it reads a file past the usual size limit.
• Claude Code's Bash and Grep cards show the description and output mode that actually ran, even when the model misnames those arguments.
• The built-in model catalog is refreshed with 8,831 models from 229 providers, up from 8,414, and now carries cache prices, so cost estimates still price cache tokens when models.dev can't be reached.

Fixed

• On Linux, built-in browser pages show inside the workspace window, on X11 and Wayland alike, instead of in a separate window. (#961, @bigFish0124; #937 reported by @lwxk)
• After Claude Code reads a large image, the rest of its reply no longer shows up twice until you reload, and neither does a reply it picks back up mid-turn, as after reaching its output limit. (#956, @nnnmdzz)
• Agents stop calling check_user_feedback over and over when you have sent no feedback, and after you answer their question they get on with the task instead of asking again whether to proceed. (#926, @Frank-zhu2000; #861 reported by @yl-3307)
• A to-do whose turn ends on an agent error, such as the upstream 400 after a cut-off tool call, now fails with the agent's own message, instead of quietly moving to To review or saying only that the agent stopped. (#945, @Frank-zhu2000; #538 reported by @pia)
• A draft switched to another agent no longer acts through the previous agent's connection before the new one is up: a canvas draft card no longer sends your message to the old agent, and a setting you change no longer lands on the old agent as its saved preference.
• Disconnecting an agent that is still starting up now stops it at once, as when you switch a draft to another agent, instead of letting it finish starting up in the background.
• Conversations open with their messages when an agent's environment variables move its data folder, such as CLAUDE_CONFIG_DIR or CODEX_HOME set in its settings, and importing finds them too, instead of opening empty. (#948 reported by @lfabay)
• A Windows desktop window connected to a codeg-server in WSL no longer stays on "Connection lost", as the connection now tries the next address when one never answers. (#930 reported by @stlyx)
• Typing no longer lags while several tiled conversations stream at once, or while an expanded sub-agent card is still running. (#935 reported by @miyling)
• Live previews of Word, Excel and PowerPoint files no longer fail when OfficeCLI takes more than 8 seconds to start, as it can on a cold start or with a large document, and quitting codeg while one is still opening no longer blocks that file's preview for the next five minutes. (#954, @ker2xu)
• Switching back to a Word, Excel or PowerPoint tab shows its live preview as you left it, instead of starting it over and reloading the document.
• On phones, the back gesture closes the file, drawer or viewer you opened, one layer at a time, instead of leaving codeg. (#940, @leehom0123)
• Question cards from Claude Code's and Qoder's own question tool show the answers you picked, multi-select ones included, instead of "No selection". (#927, @Liang-HZ)
• Qoder conversation details show the model you switched to last instead of the one the conversation started with, and Qoder's token usage now includes what its sub-agents spent, with existing conversations recounted. (#914, @Liang-HZ)
• Kimi Code conversations reopened from history read as they did live: attachments show as badges and thumbnails instead of raw <resource> text, slash commands show as you typed them, and a turn Kimi starts on its own no longer shows its internal prompt as your message.
• Kimi Code conversation titles no longer include attachment text or switch between two versions.
• Jumping to a message no longer snaps back to the bottom while a reply streams in or older messages load, whether you jump from Message navigation or from a reply back to its prompt.
• Grok's plan approvals now raise a desktop notification, as other permission requests do.
• Terminal commands an agent started are now stopped however its connection ends, also when it ends during startup or while a terminal is still being created, instead of being left running.
• On Windows and Linux, two downloads of the same link in the built-in browser no longer get mixed up when they finish.
• Switches no longer slide out of their track in Arabic.

Note

• Taskbar and Dock alerts are on by default and skip the window you are already looking at, and you can turn them off under Settings → General → Taskbar and Dock alerts.
• Cost estimates in Token Usage use base API list prices without long-context or reasoning surcharges, so they are not what a subscription plan bills, and models with no published price are left out, with the share of tokens priced shown beside the total.
• Auto-processing only picks up items opened or updated after you switch it on, unless you choose to include existing ones, and the desktop app receives webhooks only while its web service is running, which you start under Settings → Web Service.
• The web version's cache works only over HTTPS or on localhost, where browsers allow it, and the desktop app doesn't use it.
• Antigravity 1.4.0 lists one model per family, with the variant picked in a separate thinking level option, and the model you chose before the update carries over.

Thanks to @Liang-HZ, @ker2xu, @Frank-zhu2000, @leehom0123, @Adam-Dalloul, @nnnmdzz and @bigFish0124 ...

Release version 0.35.1

This one lets agents draw charts in their reply cards, and makes Generative UI a skill you enable for each agent, now named ui-render.
Alongside that: cards in replies lose their shadow, and pressing Refresh on a Skill Packs tab no longer brings back an error or an older list.

New

• Agents can draw charts in their reply cards: ask an agent with UI Render enabled to show something as a chart, and you get area, bar, line, pie or donut, radar, radial or scatter charts in your theme's colors, with values shown on hover.

Improved

• Generative UI is now enabled per agent: Settings → Skill Packs → UI lists the UI Render skill in the same agent table as the Experts tab, so you can enable it for one agent, a few or all of them.
• Cards in replies no longer cast a shadow, so they sit flat in the message like the rest of the conversation.

Fixed

• A Skill Packs tab no longer swaps its list for an error or older data when you press Refresh while it is still loading, on the Experts, Science, Office Tools and Custom tabs alike.
• Import from agent on the Custom tab of Skill Packs no longer lists codeg's own Generative UI skill, just as it already leaves out the other skill packs.

Note

• The json-render skill is now called ui-render: start a message with /ui-render ($ui-render in Codex), and if you had Generative UI on in 0.35.0, the agents that already had json-render get ui-render in its place when you upgrade.
• The "Generative UI (json-render)" switch is gone: cards in replies always render, an agent gets UI Render in conversations started after you enable it, and the cards it already drew stay after you disable it again.

发布版本 0.35.1

这一版让智能体可以在回复卡片里画图表，生成式界面也改成按智能体启用的技能，并更名为 ui-render。
同期还有：回复里的卡片去掉了阴影，技能包页面点「刷新」后不会再退回报错或旧列表。

新增

• 智能体可以在回复卡片里画图表了——让启用了「界面渲染」的智能体用图表展示，就能画出面积图、柱状图、折线图、饼图或环形图、雷达图、径向图和散点图，配色跟随主题，鼠标悬停即可查看数值。

改进

• 生成式界面改为按智能体启用——「设置 → 技能包 → 界面」里的「界面渲染」技能和「专家」页一样用表格管理，可以给单个、几个或全部智能体启用。
• 回复里的卡片去掉了阴影，和会话里的其他内容一样平铺在消息中。

修复

• 技能包页面还在加载时点「刷新」，列表不会再被换成报错或旧数据，「专家」「科学研究」「办公工具」和「自定义」页都已修复。
• 技能包「自定义」页的「从 Agent 导入」不再列出 codeg 自带的生成式界面技能，和其他技能包一样不会出现在这里。

注意

• json-render 技能更名为 ui-render——提问时用 /ui-render（Codex 里是 $ui-render）开头；如果你在 0.35.0 里打开过生成式界面，升级后已经带有 json-render 的智能体会自动换成 ui-render。
• 「启用 json-render 生成式界面」开关已移除——回复里的卡片始终会渲染；给智能体启用「界面渲染」后新开始的会话才会带上这个技能，之后再停用，它已经画出的卡片也会保留。

Release version 0.35.0

This one lets agents answer with more than text: turn on Generative UI, and a reply can show cards, tables and other interactive components right in the conversation. The chat column can be widened by dragging its edges, chat animations can be switched off, and a conversation can be maximized while files are open beside it. Claude Code moves up to 0.89.1 and Codex to 2.2.2, along with five more bundled agents.
Alongside that: the terminal panel comes back after a page reload, a workspace with many tabs opens much faster, remote workspaces keep reconnecting through outages and reopen even while their server is offline, the Open Folder dialog can create a folder, and a failed turn ends with one line saying why.

New

• Agents can reply with interactive cards and tables once you turn on "Generative UI (json-render)" under Settings → Skill Packs → UI: start a message with /json-render ($json-render in Codex) or just ask to see something as cards or a table, and the reply renders as live components while it streams.
• The chat column can be made wider or narrower by dragging its left or right edge, or under Settings → Appearance → Chat width, and double-clicking an edge puts it back to the default. (#911, @sumanx)
• Chat animations can be switched off under Settings → Appearance, so expanding, collapsing, working indicators and scrolling in the conversation happen instantly, while spinners keep turning. (#911, @sumanx)
• A conversation can be maximized while files are open beside it, from the Maximize button at the right end of its tab strip, and Restore brings the files back just as you left them.
• The Open Folder dialog can create a new folder and step into it, so a project can start from an empty folder, in the desktop app and the web version alike. (#881, @Jonathan-Asher)
• A remote workspace window can open the local workspace from Quick actions in the status bar, the macOS Window menu, or the page shown when its connection fails or expires. (#891, @ker2xu)

Improved

• Updated bundled agents: Claude Code ACP 0.89.1 (Claude Code 2.1.293), Codex ACP 2.2.2 (Codex CLI 0.160.1), Cline 3.0.70, CodeBuddy 2.164.0, Grok 1.0.51, Qoder 1.1.67, Hermes 0.21.6.
• A failed turn ends with one muted line giving the agent's own error in Codex, Claude Code, Cline and Qoder conversations, both as it happens and when you reopen the conversation.
• A workspace with many open tabs opens much faster, because a tab sitting behind another now loads its conversation the first time you switch to it instead of every tab downloading its conversation at once. (#913, @Flyneen)
• Resizing the window is smoother, as the workspace no longer runs extra full-page layouts on every step. (#898, @sumanx)
• Slash commands show the arguments they take in the / menu, the message box's Add menu and the automation editor.
• Codex answers /mcp with the live status of each MCP server, and /mcp reconnect works there too.
• Claude Code's Agent card shows the effort a sub-agent was asked to run at, next to its model.
• Claude Haiku 5.5 is now the suggested Haiku model in Claude Code's settings and in the model provider dialogs.
• On phones, the web version tints the browser bar and Android's navigation bar to match your light or dark theme, and adding it to the home screen installs codeg as an app with a full-bleed icon, even behind a sign-in proxy. (#863, @eyjohn)

Fixed

• Reloading the page or reconnecting no longer loses the terminal panel: its tabs come back with their commands still running and their output in place, and a finished command's output stays until you close its tab. (#830, @dawNotPoi; #749 reported by @kekxv)
• Claude Code no longer leaves a conversation spinning for hours after you send a message during a turn it started on its own, such as after a background task finished, and that message now gets a single reply. (#893 reported by @Luis-WowApps)
• Sending Claude Code a message while it is still writing a tool call no longer fails the turn.
• Messages you send while a Claude Code turn is running now stay in its history, and a background task that finished mid-turn shows its result there instead of looking unfinished.
• A Claude Code conversation's title no longer flips back and forth between its first and its latest message in the sidebar and in a linked chat channel topic when Claude Code can't generate a title, as behind a gateway that refuses the request.
• Replies from turns a custom agent starts on its own, such as after a background job, now stay in the conversation after a reload instead of only showing live. (#903 reported by @andrehqh)
• A custom agent that runs Claude Code's ACP adapter now takes messages inserted into the current turn directly, as the built-in Claude Code does, and those messages stay in its history. (#892 reported by @andrehqh)
• A remote workspace keeps reconnecting through network outages and shows a notice while it does, instead of giving up after three failed attempts and calling the connection expired, which now happens only when the server rejects the token. (#887, #890, @ker2xu)
• Remote workspace windows reopen at launch even while their server is offline, and pick up in place once it is back. (#887, #890, @ker2xu)
• Clicking the Dock icon on macOS brings up the workspace you used last, a minimized remote one included, instead of always the local window, and launching codeg again on Windows and Linux does the same. (#887, #889, @ker2xu)
• Custom agents from the ACP registry now install on Windows, where most binary entries looked for a doubled name like kilo.exe.exe, and an agent published as a plain executable rather than an archive can be installed too. (#905 reported by @trueai-org)
• Agents installed with their official installer into ~/.local/bin, such as Hermes, are now found when codeg isn't started from a terminal, instead of showing as not installed. (#888, @Sma1lboy; #875 reported by @toptiersnowman)
• An OpenCode 2.x installed on your system now connects, as codeg no longer passes it the --port and --hostname options its acp command rejects. (#909, @Liang-HZ)
• Effort choices now match the model you picked under Agent defaults in Settings → Collaboration and in the to-do and automation editors, instead of following the agent's default model. (#910, @Liang-HZ)
• Scrolling up while a reply streams no longer pulls you back to the bottom, whether you use the wheel over a code block, touch, the keyboard or the scrollbar. (#899, @Adam-Dalloul)
• Reordering conversation tabs no longer leaves a tab's transcript blank until you scroll it. (#901, @Adam-Dalloul)
• In a sent message, a /word or $word shows as a command only when the agent actually offers that command or skill, so paths like /tmp stay plain text. (#902, @Adam-Dalloul)
• The tab strip no longer squeezes and springs back when you open or close a side panel. (#912, @sumanx)
• The sidebar, side panel and terminal panel no longer wobble while you resize the window. (#897, @sumanx)
• Collapsing a thinking block or tool card no longer flashes its content for a frame, and after a click in the conversation, pressing Esc, copying or coming back from another window no longer draws a focus outline around it. (#896, @sumanx)
• A conversation that fails to load now retries at growing intervals, up to every 30 seconds, with the error and Reload button staying in view, instead of re-requesting nonstop while the error flickers.
• A conversation no longer stays stuck loading when a background sync or a page of older messages arrives while it loads.
• Token usage is counted more accurately: Codex no longer counts cache writes as fresh input, with existing Codex conversations recounted, and CodeBuddy totals now include requests that end in a tool call.
• Cline history no longer shows the messages Cline marks as internal, such as compaction summaries, hook context and the prompt it sends itself after you approve switching to Act mode.
• Claude Code conversations reopened from history show a 1M context window for models that have one natively, such as Opus 4.7 and later, Sonnet 5 and later, Haiku 5.5, Fable and Mythos, instead of 200K.
• Backups now include Codex's conversation names, and restoring merges Codex's and Kimi's session lists with what is already there instead of skipping or overwriting them, so restored conversations keep their names and folders.
• On macOS, computer use presses return after typing into a Chromium or Electron app when the agent asks to submit, instead of treating the typing as failed because the field reads back empty.
• The desktop app writes its logs to the same folder however it is started, and Settings → Runtime Logs opens the folder actually in use and shows each file's current size. (#893 reported by @Luis-WowApps)

Note

• Codex ACP 2.2 needs Node.js 20.3.0 or later, as every message fails on Node.js 20.0 through 20.2, and the environment diagnostics now check the full Node.js version each agent needs rather than only the major version.
• With Claude Code ACP 0.88.0 and later, a running Monitor no longer shows in the background task strip, so it can't be stopped from there.
• Generative UI is off by default: agents get the json-render skill in conversations started after you turn it on, turning it off removes the skill from every agent, and a json-render skill of your own is never replaced.
• A tab restored behind another shows a brief loading state the first time you switch to it, while its conversation loads.
• The terminal panel comes back after a page reload but not after codeg itself restarts, since a restart ends the terminals' processes and their tabs then report the session as unavailable.

Thanks to @sumanx, @ker2xu, @Adam-Dalloul, @Liang-HZ, @dawNotPoi, @e...

Release version 0.34.0

This one lets you tag conversations: colored tags for every conversation or for one folder, shown after the title and ready to filter the sidebar by. The built-in browser can show a page as a tablet, a phone or any size you type, and Claude Code moves up to 0.86.0 along with six more bundled agents.
Alongside that: codeg's GitHub home is now spacering-net/codeg, a long Claude Code conversation no longer runs codeg out of memory, Claude Code answers /mcp in the conversation, and a queued message you insert as a turn ends goes out first.

New

• You can tag conversations from the conversation header or the Tags item in a sidebar conversation's right-click menu, and create, edit, reorder or delete tags in the "Manage tags" dialog.
• A tag is either global or a folder's own: global tags are offered on every conversation, chat mode included, and a folder's tags only on conversations in that folder and its worktrees.
• The sidebar can be filtered by tags, showing the conversations that have any or all of the tags you pick.
• A branch tag can show the git branch each conversation started on after its title, in a color you choose, once you turn it on in "Manage tags".
• The built-in browser can show a page as a tablet, a phone or at a width and height you type, from the device button to the right of the address bar, and each tab keeps its choice even after a restart.

Improved

• Updated bundled agents: Claude Code ACP 0.86.0 (Claude Code 2.1.287), OpenCode 1.18.35, Cline 3.0.68, CodeBuddy 2.161.4, Antigravity 1.3.0, OpenClaw 2026.9.8, Cursor 2026.10.01.
• Claude Code answers /mcp in the conversation with a list of your MCP servers, and /mcp reconnect, enable and disable work there too.
• Claude Code shows its Agent option again when you have custom Claude Code subagents, so a conversation can run as one of them from the next message on.
• Antigravity conversations show their context window usage in the status bar, which Antigravity 1.3.0 now reports.
• The built-in model catalog is refreshed with 8,414 models from 226 providers, up from 8,339 models.

Fixed

• codeg no longer runs out of memory when Claude Code writes a large batch to a conversation's transcript at once, as it can when picking a long conversation back up, which used to push codeg past 13 GB until the system killed it. (#885 reported by @andrehqh)
• Claude Code keeps your Fable pick after the update to 0.86.0, which renamed the model, instead of switching to the default model every time a conversation connects.
• A queued message you click "Insert into current turn" on goes out first if the turn ends before it gets in, with a notice saying so, instead of waiting behind the messages queued before it. (#883, @leehom0123)
• Edits you make in the message box while a message is being inserted into the current turn are kept instead of being wiped afterwards, and clicking insert on queued messages in quick succession no longer inserts two at once. (#886, @leehom0123)
• A CodeBuddy sub-agent's tool calls stay inside its Agent card while they are still being written, with CodeBuddy 2.161.4.

Note

• codeg now lives at github.com/spacering-net/codeg, and old links, update checks from installed versions and the install scripts are redirected there, so nothing needs to change.
• Docker images now go to ghcr.io/spacering-net/codeg, while ghcr.io/xintaofei/codeg stays at 0.33.0, so switch to the new address to keep updating; xintaofei/codeg on Docker Hub carries on as before.
• The browser's tablet, phone and custom views only change the page size — sites still see a desktop browser with mouse input, so one that picks its mobile version by browser type keeps showing the desktop one.
• With Antigravity 1.3.0, a "Bypass proxy for" list of just * no longer keeps all of its traffic direct — when a proxy is configured, some of its requests go through the proxy again.

Thanks to @leehom0123 for contributing to this release, and to @andrehqh for the report.

发布版本 0.34.0

这一版可以给会话打标签了：彩色标签可以所有会话通用，也可以只属于某个文件夹，显示在标题后面，侧边栏也能按标签筛选。内置浏览器能把网页显示成平板、手机或你输入的任意尺寸，Claude Code 升级到 0.86.0，另有六个内置智能体更新了版本。
同期还有：codeg 的 GitHub 仓库迁到了 spacering-net/codeg，很长的 Claude Code 会话不再让 codeg 耗尽内存，Claude Code 能在会话里回应 /mcp，回合结束时插入的排队消息也会最先发出。

新增

• 可以给会话打标签了——在会话头部或侧边栏会话右键菜单的「标签」里添加，标签的新建、编辑、排序和删除都在「管理标签」对话框里完成。
• 标签分为全局标签和文件夹标签——全局标签所有会话都能用，聊天模式也不例外；文件夹标签只出现在该文件夹及其工作树的会话里。
• 侧边栏可以按标签筛选，显示带有任一或全部所选标签的会话。
• 分支标签可以把会话创建时所在的 git 分支显示在标题后面，颜色可自选，在「管理标签」里打开即可。
• 内置浏览器能把网页显示成平板、手机或你输入的宽高——在地址栏右侧的设备按钮里切换，每个标签页都会记住自己的选择，重启后也不变。

改进

• 内置智能体版本更新： Claude Code ACP 0.86.0（Claude Code 2.1.287）、OpenCode 1.18.35、Cline 3.0.68、CodeBuddy 2.161.4、Antigravity 1.3.0、OpenClaw 2026.9.8、Cursor 2026.10.01。
• Claude Code 能在会话里回应 /mcp 了，会列出你的 MCP 服务器，/mcp reconnect、enable、disable 也能直接在会话里用。
• 有自定义的 Claude Code 子智能体时，「Agent」选项又回来了，选中后会话从下一条消息起按该子智能体的设定运行。
• Antigravity 会话会在状态栏显示上下文窗口用量，这是 Antigravity 1.3.0 新提供的数据。
• 内置模型目录已更新，共 226 个供应商、8414 个模型，此前为 8339 个。

修复

• Claude Code 一次往会话记录文件写入大量内容时，codeg 不再耗尽内存——续接很长的会话时就可能这样写入，以前会让 codeg 涨到 13 GB 以上，直到被系统强行结束。（#885 由 @andrehqh 反馈）
• Claude Code 升级到 0.86.0 后，你选的 Fable 仍然有效，不会因为新版改了 Fable 的模型标识而在每次连接时退回默认模型。
• 在排队消息上点了「插入当前回合」、但回合在插入前就结束时，这条消息会排到最前、作为下一回合最先发出，并给出提示，不再排在它前面的消息后面。（#883，@leehom0123）
• 消息正在插入当前回合时，你在输入框里接着做的修改会保留下来，不再被清空，在排队消息上快速连点插入也不会同时插进两条。（#886，@leehom0123）
• CodeBuddy 子智能体的工具调用在参数生成过程中也会留在它的 Agent 卡片里，需要 CodeBuddy 2.161.4。

注意

• codeg 的 GitHub 仓库已迁到 github.com/spacering-net/codeg，旧链接、已安装版本的更新检查和安装脚本都会自动跳转过去，无需任何改动。
• Docker 镜像改为发布到 ghcr.io/spacering-net/codeg，ghcr.io/xintaofei/codeg 停在 0.33.0，想继续更新请换成新地址；Docker Hub 上的 xintaofei/codeg 照常更新。
• 内置浏览器的平板、手机和自定义视图只改变页面尺寸——网站看到的仍是桌面浏览器和鼠标操作，按浏览器类型切换手机版的网站仍会显示桌面版。
• Antigravity 1.3.0 下，「不走代理的地址」只填 * 已不能让它的流量全部直连——配置了代理时，它有一部分请求会重新走代理。

感谢 @leehom0123 为本次发布做出的贡献，也感谢 @andrehqh 的反馈。

Release version 0.33.0

This one lets agents see and operate your desktop. With the new Computer Use preview, you share a window, a whole app or the entire screen, and agents can read it and click, type and drag in it, with Stop sharing always one click away. Codex moves up to its 2.1.1 adapter and Claude Code to 0.85.1, and seven more bundled agents get new versions.
Alongside that: the desktop app reopens the workspace windows you had open when you quit, the on-screen keyboard stays down when you switch conversations on a touch device, icons draw a little lighter, a to-do runs the agent its picker shows, and on Windows codeg-server installs into a folder of its own.

New

• Agents can see and operate the windows you share with them (preview) — switch it on under Settings → Computer Use, share a window from the status bar's Computer use panel as "Read only" or "Read and act", and agents can take screenshots, read what the window shows, and click, scroll, type and press keys in it on macOS, Windows and Linux, while the panel lists what they just did; nothing is readable until you share it. (#870)
• One click or one shortcut stops everything — "Stop sharing" in the Computer use panel or on the floating stop bar, or Ctrl+⌘+Esc (Ctrl+Alt+Esc on Windows and Linux), ends every sharing and cuts off whatever agents are doing, and a shared window no agent has used for 30 minutes (adjustable) stops being shared by itself. (#870)
• codeg's own windows and your never-share list stay off limits — password managers and system settings are on the list by default, and you can add or remove apps. (#870)
• You can share a whole app or the entire screen — an app shared as a whole includes the windows it opens later, its menus and its own shortcuts, and "Offer the entire screen" (macOS and Windows, off by default) lets agents see one picture of the screen and click anywhere on it, with codeg's windows and never-share apps painted over. (#870)
• Agents can do more than click — drag, hold keys, restore minimized or hidden windows and, for apps that take no keys in the background (such as Edge, Chrome and VS Code on Windows), bring the window to the front, which you can switch off; opening apps, moving windows and using the clipboard have their own switches, off by default, and an agent can only paste what it put on the clipboard itself. (#870)
• codeg-server can offer computer use too — start it with CODEG_COMPUTER_USE=1 and its web clients can share the server machine's windows and stop them from the same panel. (#870)
• The desktop app reopens the workspace windows you had open when you quit — the local one and each remote workspace, with the one you were using back in front, and a remote workspace that can't be reached is reported in the local window.

Improved

• Updated bundled agents: Claude Code ACP 0.85.1 (Claude Code 2.1.286), Codex ACP 2.1.1 (Codex CLI 0.159.3), OpenCode 1.18.34, Cline 3.0.67, CodeBuddy 2.161.0, Grok 1.0.46, Qoder 1.1.65, OpenClaw 2026.9.7, Cursor 2026.09.28.
• Forking a conversation no longer leaves the original session held by the agent — codeg closes it after the fork, so Codex lets go of it within about a minute and Claude Code stops keeping a process running for every fork, and the "open elsewhere" banner now says to reload in about a minute if you just forked.
• Icons draw a little lighter — those at the icon set's default weight of 2 now use 1.75, which sits better beside small text, while icons given a deliberate weight keep it. (#866, @SousekiL)
• Settings → "Browser" is now "Browser Use", next to the new "Computer Use" page, and "Collaboration" moves up to follow "General". (#870)
• On Windows, codeg-server installs into its own folder, %LOCALAPPDATA%\codeg-server, so neither install overwrites the other's files or stops the other's processes, and re-running the install script moves a server that shares the desktop app's folder out of it. (#870)
• The desktop app drops about 70 MB it never used — the standalone server binary is no longer bundled with it. (#870)
• The built-in model catalog is refreshed — 8,339 models from 225 providers, up from 8,281 models.

Fixed

• On a touch device, switching to another conversation no longer pops up the on-screen keyboard over it — tapping the message box still brings it up, and desktop behavior is unchanged. (#865, @Flyneen)
• A to-do runs the agent its picker shows — on a fresh install, or when the usual agent is disabled, the picker could show an agent that wasn't actually saved, and the to-do then failed to start with "no agent configured"; now the agent shown is saved with the to-do, and the hint under the picker says whether it is inherited or saved. (#864 reported by @qq974969638)
• Quitting with ⌘Q, from the Dock or by logging out now runs the full quit cleanup — agent processes, terminals and the web service are stopped as they are when you quit from inside codeg, and logging off on Windows does the same.
• Cline runs on macOS 27 — macOS 27 killed Cline 3.0.65 and earlier at launch, while 3.0.67 ships with a valid signature, so upgrade it in Settings → Agents; OpenCode 1.18.34 is now properly signed too.
• A background command in a Claude Code conversation no longer has its card overwritten with internal marker text when it finishes, and one that moved to the background because you sent a message or it timed out shows the Background badge.
• Codex conversations reopen more faithfully — a question you answered with your own text under "Other" comes back as that text rather than as two picks, a failed compaction shows the service's error message instead of raw JSON, and a conversation started in the Codex desktop app shows its attachments as links followed by your request, titled after what you wrote.
• A search pipeline that finds nothing no longer shows as a failed command — a command like rg --files src | rg foo that ends in rg or grep and matches nothing now reads as completed with its exit code still shown, instead of a red failure counted as an error in its tool group, while real errors such as exit code 2 or an error message still show as failed. (#877 reported by @Ryn-Mic)
• Grok's model picker picks up the full model list — in a session opened before Grok had fetched its model catalog, such as after your sign-in expired, the picker kept the short built-in list for the rest of the connection; it now updates as soon as Grok announces the catalog, and the model and reasoning effort you chose stay selected. (#876 reported by @goon-13)

Note

• Computer use downloads the open-source cua-driver (0.32.0) the first time it is needed, or when you press Install in Settings → Computer Use.
• On macOS, grant Accessibility and Screen Recording to codeg-computer-helper, not to codeg — only codeg can use that helper, so agents' own shells can't borrow the permissions; Windows and Linux have no such separation, so there the settings decide what agents get through codeg.
• On Windows, codeg-server's web files now live in %LOCALAPPDATA%\codeg-server\web — if your server moves there, update a CODEG_STATIC_DIR that still points at %LOCALAPPDATA%\codeg\web.
• With Claude Code 2.1.286, sending a message moves a running command, MCP call or sub-agent to the background instead of waiting for it or interrupting it — codeg marks such a command with the Background badge, and a question card still open is dropped when you send, as before.
• Behind a custom API address, Claude Code's model list no longer has separate "(1M context)" entries — since 2.1.285 those models use their 1M window by default, and a conversation saved on such an entry carries on with the plain model; if your gateway stops at 200K, Claude Code's advice is /autocompact 200k.
• Cline 3.0.66 and later count output tokens without reasoning tokens, so reasoning models show lower output token counts.

Thanks to @Flyneen and @SousekiL for contributing to this release, and to @qq974969638, @Ryn-Mic and @goon-13 for the reports.

发布版本 0.33.0

这一版让智能体能看见并操作你的桌面：在新的「电脑操作」（预览）里，你可以共享一个窗口、一个应用或整个屏幕，智能体就能读取画面，并在其中点击、输入、拖拽，「停止共享」则随时一键就能按下。Codex 升级到 2.1.1 适配器，Claude Code 升级到 0.85.1，另有七个内置智能体更新了版本。
同期还有：桌面版会重新打开你退出时开着的工作区窗口，触屏设备切换会话时不再弹出软键盘，图标线条细了一点，待办任务会运行选择器里显示的智能体，Windows 上的 codeg-server 也装进了自己的文件夹。

新增

• 智能体可以查看并操作你共享给它的窗口（预览）——在「设置 → 电脑操作」里启用后，从状态栏的「电脑操作」面板把窗口共享为「仅读取」或「读取并操作」，智能体就能在 macOS、Windows 和 Linux 上截图、读取窗口内容，并在窗口里点击、滚动、输入和按键，面板里还会列出它们刚做了什么；没有共享的窗口一律读不到。（#870）
• 一键或一个快捷键就能停下全部——「电脑操作」面板或悬浮停止条上的「停止共享」，以及快捷键 Ctrl+⌘+Esc（Windows、Linux 上是 Ctrl+Alt+Esc），都会立即结束所有共享并中断智能体正在做的事；共享的窗口 30 分钟没有智能体用过（时长可调）也会自动停止共享。（#870）
• codeg 自己的窗口和「永不共享」名单里的应用始终无法共享——名单默认包含密码管理器、系统设置等，可以自行增删。（#870）
• 可以共享整个应用或整个屏幕——共享整个应用时，它之后打开的窗口、菜单和自己的快捷键都包含在内；「允许共享整个屏幕」（仅 macOS 和 Windows，默认关闭）开启后，智能体能看到整屏画面并在任意位置点击，codeg 自己的窗口和永不共享的应用会被遮盖。（#870）
• 智能体能做的不止点击——还能拖拽、按住按键、恢复被最小化或隐藏的窗口，遇到在后台收不到按键的应用（比如 Windows 上的 Edge、Chrome、VS Code），还能把窗口切到前台（可关闭）；打开应用、移动窗口和使用剪贴板各有单独的开关，默认关闭，智能体也只能粘贴它自己放进剪贴板的内容。（#870）
• codeg-server 也可以提供电脑操作——启动时设置 CODEG_COMPUTER_USE=1，网页端就能在同一个面板里共享服务器所在机器的窗口，也能随时停止共享。（#870）
• 桌面版会重新打开你退出时开着的工作区窗口——本地的和每个远程工作区都一样，退出时正在用的那个回到最前面；连不上的远程工作区会在本地窗口里提示。

改进

• 内置智能体版本更新： Claude Code ACP 0.85.1（Claude Code 2.1.286）、Codex ACP 2.1.1（Codex CLI 0.159.3）、OpenCode 1.18.34、Cline 3.0.67、CodeBuddy 2.161.0、Grok 1.0.46、Qoder 1.1.65、OpenClaw 2026.9.7、Cursor 2026.09.28。
• 分叉会话后，原会话不再被智能体一直占着——分叉之后 codeg 会关掉原会话，Codex 约一分钟内就会放开它，Claude Code 也不再为每次分叉多留一个进程；「会话在别处打开」的提示现在也会说明：刚分叉过的话，约一分钟后重新加载即可。
• 图标线条略微变细——原本使用图标库默认粗细 2 的图标改为 1.75，与小号文字搭配更协调，特意指定过粗细的图标保持不变。（#866，@SousekiL）
• 「设置」里的「浏览器」更名为「浏览器操作」，与新增的「电脑操作」并排，「协作」也挪到了「常规」后面。（#870）
• Windows 上的 codeg-server 装进自己的文件夹 %LOCALAPPDATA%\codeg-server，两边安装时互不覆盖文件，也不再互相结束对方的进程；重新运行安装脚本，会把与桌面版共用目录的服务器挪出来。（#870）
• 桌面版去掉了约 70 MB 用不上的内容——不再捆绑独立服务器程序。（#870）
• *内置模型目录已更新...

chore(release): rehearse 0.33.0-rc.2 from the computer-use branch

A release candidate built from task/227 (PR #870) before it is merged.
Since rc.1: on Windows codeg-server installs into a folder of its own,
apart from the desktop app's, and each installer stops only its own
processes; computer use runs cua-driver 0.32.0, adapted to its element
tokens and its one snapshot per window. It is not on main and is not
meant for general use.

发布流水线预演：电脑操作分支（task/227，PR #870）合入前构建的候选版本。相对 rc.1：Windows 上 codeg-server 装到自己的目录，不再和桌面版共用，两边的安装器只结束自己的进程；电脑操作改用 cua-driver 0.32.0，适配了它的元素令牌和每个窗口只留一个快照的新模型。不在 main 上，不供日常使用。