需 Node.js 20.6 及以上版本；安装后要执行 webcmd skills add 并选择 Claude、Codex 等 harness 才会装上 webcmd-browser 技能；原始浏览器命令必须显式指定可读的 Session ID，并行 Agent 应各自建独立 Session
webcmd
agentrhqApache-2.0活跃维护用命令行驱动浏览器的自学习 Agent，基于 Playwright 自动操作网页，并在使用中积累经验、越用越顺手，适合做网页自动化与数据采集。
Self-learning agent browser
webcmd-v0.8.4发布于 2026-09-09 04:36 UTC2,768star1,537fork同步于 2026-11-10 11:43 UTC
npm install -g @agentrhq/webcmd
webcmd skills add 硬件要求按 README 说明
README 未说明；仅要求 Node.js 20.6+，浏览器运行在本地或 Webcmd Cloud 托管基础设施上
给 AI Agent 用的自学习浏览器基础设施，把网站导航经验存成本地记忆，省去重复探索
用命令行驱动浏览器的自学习 Agent，基于 Playwright 自动操作网页，并在使用中积累经验、越用越顺手，适合做网页自动化与数据采集。
核心能力
- 用 webcmd browser 在真实浏览器里检查、点击、输入、提取并抓取网络请求
- 把观察到的页面、状态、动作、工作流和 API 沉淀为 Agent 可读的站点地图记忆
- 用 Profile 管理登录态 cookie，用 Session 隔离并行 Agent 的浏览器窗口
- 支持向指定会话提交 Playwright 风格脚本做多步探索
- 记忆失败不阻塞任务，首次访问可用云端种子，后续学习留在本地
容易踩的坑
内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README
- webcmd-v0.8.4webcmd: v0.8.4最新正式版2026-09-09 · 2026-09-09 04:36 UTC
Improvements
webcmd setupnow offers interactive Chrome profile syncing when Chrome is selected. The new prompt defaults to disabled and does not appear when the option was already specified.
Fixes
webcmd profile listnow shows saved but disconnected profiles in plain-text output, even when no runtime profiles are currently connected.- Chrome profile exports now run shortly after command activity, even when tabs remain open. Rapid commands are debounced into a single export without closing browsers or pages.
- Exported Chrome profiles now use their webcmd alias as the display name when registration succeeds. This naming update is best-effort when Chrome is already running.
Contributors
- webcmd-v0.8.3webcmd: v0.8.32026-09-08 · 2026-09-08 17:29 UTC
查看更新说明
Highlights
- Added
webcmd setup --browser chrome --sync-to-chrometo automatically copy profile cookies into a webcmd-managed native Chrome profile when the browser runtime closes. The profile is registered with Chrome in the background and can then be opened directly in Chrome. - Added persistent, versioned agent behavior settings for each native SLAB profile, providing consistent humanized browser interactions across daemon sessions.
Improvements
- Humanized input now avoids intentional mistypes in sensitive fields such as passwords, payment fields, and one-time-code inputs.
- Disposing an agent page now cancels in-progress humanized input and delays immediately.
- Playwright tracing snapshots are disabled on agent-controlled pages.
Fixes
- Browser network-capture waits now use monotonic elapsed time, preventing premature timeouts when the system clock changes.
Contributors
- Added
- webcmd-v0.8.2webcmd: v0.8.22026-09-07 · 2026-09-07 14:40 UTC
查看更新说明
Improvements
webcmd setupnow exclusively configures local browser mode. The hosted/local prompt and the--modeand--api-keyoptions have been removed. Browser selection, Chrome cookie import, status reporting, and daemon restart behavior remain available.- On first access to a product without local site memory, Webcmd now requests a seed from
https://api.webcmd.devby default. The unauthenticated request sends only the resolved product/domain, uses a two-second timeout, and is not retried. SetWEBCMD_GLOBAL_MEMORY=offto disable lookups;WEBCMD_GLOBAL_MEMORY_URLremains available as a developer or test override.
Fixes
webcmd skills removenow removes skills from only one provider and scope—or one custom--path—instead of scanning every supported location. It supports--providerand--scope, prompts for them interactively when needed, and includes them in JSON output. Stable links under~/.webcmd/skillsare no longer removed by this command.webcmd skills addandwebcmd skills updatenow prune stale symlinks from~/.webcmd/skillswhile preserving user-owned entries that are not symlinks.
Contributors
- webcmd-v0.8.1webcmd: v0.8.12026-09-04 · 2026-09-04 15:02 UTC
查看更新说明
Highlights
- Local Chrome setup can now import cookies from installed Chrome profiles, allowing Webcmd to start with existing sign-ins. Setup maps Chrome’s default profile to Webcmd’s
defaultprofile and creates stable names for additional profiles. - Use
--chrome-profileto import one Chrome profile, or--import-chrome-cookiesand--no-import-chrome-cookiesto control cookie import in non-interactive setup.
Improvements
- Redesigned
webcmd --helpwith an ASCII banner, clearer command groups, global options, examples, agent guidance, and dynamic site/external CLI lists. Structured JSON and YAML help remain unchanged. - Browser readiness messages now identify the configured runtime and provide relevant guidance for SLAB, Cloak, Chrome, and custom runtimes.
Fixes
- Improved SLAB session orchestration:
- Native profile IDs such as
Profile 1are now supported. session listexposes unbound SLAB windows as temporarydiscoveredsessions.browser tabsdiscovers pages without focusing or claiming them.browser bind --page <page-id>claims the page’s complete window, including sibling tabs.- Closing an adopted human session detaches Webcmd while leaving the window open.
browser close --page <page-id>can close an exact tab;--forcepermits closing adopted human tabs.- Explicit session closure now cleans up SLAB session records correctly.
- The initial blank page is claimed instead of leaving an unowned tab behind.
- Native profile IDs such as
- Fixed installed Chrome on macOS unloading its profile or terminating after the last visible probe window closes.
- Browser daemon startup no longer inherits caller-owned temporary-directory environment variables, improving reliability when launched from short-lived environments.
Contributors
- Local Chrome setup can now import cookies from installed Chrome profiles, allowing Webcmd to start with existing sign-ins. Setup maps Chrome’s default profile to Webcmd’s
- webcmd-v0.8.0webcmd v0.8.0: Lean, Mean and Modular2026-09-02 · 2026-09-02 09:52 UTC
查看更新说明
Highlights
- Local self-learning site memory: Browser workflows can now retain verified navigation and workflow knowledge under
~/.webcmd/sites. Learning stays out of normal command output and includes candidate capture and search, conflict-aware checkpointing, optional seed initialization, and bounded memory files. - Selectable local browser runtimes: Keep bundled Cloak as the default, reuse an installed Google Chrome, opt into SLAB on macOS, or select a compatible Chromium executable by absolute path.
- One browser skill contract:
webcmd skills addand supported plugins now install a singlewebcmd-browserskill. Adapter, search, fetch, and plugin workflows remain available directly through the CLI.
Improvements
- Installed Google Chrome uses an isolated
~/.webcmd/chrome/profilesdirectory and a Webcmd-owned, nonzero loopback CDP port. Normal headed launches retain Chrome’s nativenavigator.webdrivervalue, without claiming general automation undetectability. - Custom Chromium builds receive separate, deterministic profile directories, preventing them from overwriting managed Cloak profiles.
webcmd setup --statusnow reports the configured browser selection, whilewebcmd doctoridentifies both the selected browser and live runtime.- SLAB setup reuses an existing installation and verifies its code signature and bundle identifier before launch.
- Chrome process ownership, cleanup, locked-profile recovery, and CDP endpoint verification have been hardened.
- Browser page binding can use either a stable page selector or target ID.
- Optional site-memory seeds are fetched only when
WEBCMD_GLOBAL_MEMORY_URLis configured and no local memory exists. SetWEBCMD_GLOBAL_MEMORY=offto disable seed lookup. - Candidate evidence can record the public egress IP locally; set
WEBCMD_CANDIDATE_PUBLIC_IP=offto disable that lookup. Candidate provenance is not included in ordinary output or pushed by the local site-memory repository. - Added setup guides for Aider and Cline, and refined agent-harness navigation and naming across the documentation.
If you used beta site memory, remove the affected product’s old
~/.webcmd/sites/<product>/sitemap/SITE.mdbefore first use, or remove that product directory if it is no longer needed. Beta memory is not guaranteed to migrate.Contributors
- Local self-learning site memory: Browser workflows can now retain verified navigation and workflow knowledge under
用户点评
登录用户的真实使用记录，可按版本和硬件筛选共 0 条
还没有人点评。第一个说出真实体验的人最有价值。
Webcmd
Self-learning browser infra for AI agents.
Webcmd learns the navigational context of websites as agents use them, then turns that knowledge into local memory for faster, cheaper, more reliable browser automation. The goal is simple: stop making agents rediscover the same sites on every run and cut browser-agent token spend by up to 90%.
Webcmd pairs live browser control with a self-learning memory layer:
| Layer | Scenario | What Webcmd Helps With |
|---|---|---|
| 0. Live browser control | The site is unfamiliar. | Use webcmd browser to inspect, click, type, extract, capture network calls, and complete the task in a real browser. |
| 1. Sitemap memory | The site is familiar, but the action space is not fully known. | Capture an agent-facing sitemap of observed pages, states, actions, workflows, APIs, pitfalls, and fallback paths. |
Demo
https://github.com/user-attachments/assets/bdb65307-9e2a-4d58-9175-45d59528ae37
Proxy Partner Giveaway (courtesy: Swiftproxy)
Power your web automation and data collection with Swiftproxy, a residential proxy provider offering rotating and static residential proxies for reliable web access. With 90M+ residential IPs across 220+ locations, Swiftproxy helps support automated browsing and web-based workflows. Get 10% off with code PROXY90.
Link: https://www.swiftproxy.net/?ref=webcmd
Quick Start
Agent prompt
Fetch and follow https://raw.githubusercontent.com/agentrhq/webcmd/main/start.md to set up Webcmd end to end.
Manual
Webcmd requires Node.js 20.6+.
npm install -g @agentrhq/webcmd
webcmd skills add
When prompted, choose Claude, Codex, another supported harness, or a custom
skills path. That installs exactly one skill, webcmd-browser.
Load or tag webcmd-browser only for live browser work, then describe the
outcome you want. Installation and setup commands do not require that skill.
Use webcmd to research the latest discussions about browser automation across Hacker News and Reddit, then return a concise comparison with source links.
What You Can Ask
- “Use webcmd to research agentic browser automation on PubMed and return the title, authors, publication date, abstract, and URL for each result.”
- “Use webcmd to find active AI infrastructure companies in the YC company directory and return the company, batch, description, location, profile URL, and source links. Keep it read-only.”
- “Use webcmd to look up parts on Grainger by part number and return price, stock, minimum order quantity, lead time, and product URL.”
- “Use webcmd with my logged-in
workprofile to summarize unread LinkedIn messages from the last seven days and return the sender, subject or opening text, received time, and conversation URL.” - “Use webcmd to check Grainger part prices and SAP Ariba purchase-order status, then return a combined summary.”
See It in Action
Use webcmd with my logged-in `social` profile to collect my recent X bookmarks and return the author, text, and URL.
The agent uses the logged-in profile to complete the task in a real browser. Along the way, Webcmd quietly retains useful navigation context so later agents can avoid repeating the same exploration.
Where Webcmd Works
Webcmd can work through authenticated browser sessions across research, social, AI, shopping, and booking products.
| Group | Supported surfaces | Representative outcomes |
|---|---|---|
| research and communities | Hacker News, Reddit, PubMed | Compare current discussions, find primary research, and return concise summaries with source links. |
| social and professional | X/Twitter, LinkedIn, TikTok | Collect bookmarks, monitor public posts, or research people and creators with a named profile when needed. |
| AI tools | ChatGPT, Claude, Gemini, NotebookLM | Retrieve conversations, research outputs, notebooks, and generated materials from the tools you already use. |
| shopping and bookings | Amazon, Blinkit, Zepto, BigBasket, District, Practo | Compare products, availability, prices, appointments, events, and delivery options. |
This list is illustrative. Webcmd can operate other websites through the same live browser workflow.
How Self-Learning Works
Learning stays quiet and selective: the live browser is always truth, Webcmd never explores just to learn, and a memory failure never blocks the task. First access may use a Webcmd Cloud seed; subsequent learning stays local.
For local, multi-step browser exploration, agents can send one sandboxed Playwright-style program to an explicit browser session:
webcmd --profile work session create "Work Project" -f json
# id: work-project-k7
webcmd --profile work --session work-project-k7 browser tabs
webcmd --profile work --session work-project-k7 browser run --file explore.js
printf 'return await page.title();' \
| webcmd --profile work --session work-project-k7 browser run --stdin
webcmd --profile work session close work-project-k7
Profiles are cookie jars; Sessions are independent browser windows within a profile, so Session IDs are immutable, Profile-scoped, and safe to reuse for that Session's lifetime. Parallel agents should create separate Sessions. Raw browser commands require an explicit readable Session ID.
Benchmarks
On BU Bench V1, a 100-task browser automation benchmark, Webcmd recorded the highest accuracy and lowest estimated controller cost per completed task, and fewest agent turns per completed task in this comparison.
All tools used the same Pi controller, controller model, Codex gpt-5.4 judge,
and CloakBrowser engine. This is a stronger judge than the original BU Bench
setup, whose current runner uses Gemini 2.5 Flash.
Accuracy is passed tasks out of 100. Cost and agent turns are averaged over
completed tasks; cost excludes judge usage. See the
benchmark report for category results, methodology,
architectural analysis, and reproduction steps.
Learn More
Webcmd Cloud can run supported commands and browser sessions on hosted infrastructure. It is in active development and is not yet stable.
Contributing
See CONTRIBUTING.md.
License
Released under the terms in LICENSE.