Browser Use
Browser agents and stealth cloud browsers for AI: run web tasks, fill forms, extract data.
作者提供的说明(英文原文)
Jira and Confluence behind SSO via your own browser session: every mcp-atlassian tool, no API token
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
JIRA_URL, CONFLUENCE_URL, READ_ONLY_MODE, ENABLED_TOOLS, ATLASSIAN_BROWSER_AUTH_ENABLED, ATLASSIAN_STORAGE_STATE你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"atlassian-browser": {
"command": "uvx",
"args": [
"atlassian-browser-mcp"
],
"env": {
"JIRA_URL": "your-value",
"CONFLUENCE_URL": "your-value",
"READ_ONLY_MODE": "your-value",
"ENABLED_TOOLS": "your-value",
"ATLASSIAN_BROWSER_AUTH_ENABLED": "your-value",
"ATLASSIAN_STORAGE_STATE": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add atlassian-browser --env JIRA_URL=your-value --env CONFLUENCE_URL=your-value --env READ_ONLY_MODE=your-value --env ENABLED_TOOLS=your-value --env ATLASSIAN_BROWSER_AUTH_ENABLED=your-value --env ATLASSIAN_STORAGE_STATE=your-value -- uvx atlassian-browser-mcp运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"atlassian-browser": {
"command": "uvx",
"args": [
"atlassian-browser-mcp"
],
"env": {
"JIRA_URL": "your-value",
"CONFLUENCE_URL": "your-value",
"READ_ONLY_MODE": "your-value",
"ENABLED_TOOLS": "your-value",
"ATLASSIAN_BROWSER_AUTH_ENABLED": "your-value",
"ATLASSIAN_STORAGE_STATE": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"atlassian-browser": {
"type": "stdio",
"command": "uvx",
"args": [
"atlassian-browser-mcp"
],
"env": {
"JIRA_URL": "your-value",
"CONFLUENCE_URL": "your-value",
"READ_ONLY_MODE": "your-value",
"ENABLED_TOOLS": "your-value",
"ATLASSIAN_BROWSER_AUTH_ENABLED": "your-value",
"ATLASSIAN_STORAGE_STATE": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
JIRA_URL envCONFLUENCE_URL envREAD_ONLY_MODE envENABLED_TOOLS envATLASSIAN_BROWSER_AUTH_ENABLED envATLASSIAN_STORAGE_STATE env这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月7日
注册名称已通过 GitHub 账户 GeiserX 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(1.4.1)于10天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Python (uv)。
同一分类中有:Browser Use, Chrome DevTools MCP, MCP Server Browser
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Browser agents and stealth cloud browsers for AI: run web tasks, fill forms, extract data.
MCP server for Chrome DevTools
MCP server for browser use access
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
MCP server that wraps the upstream mcp-atlassian toolset with browser-cookie authentication via Playwright, for Atlassian Server/Data Center instances behind corporate SSO (Okta, SAML, etc.) where API tokens are not available. It runs locally over stdio: you log in once with the bundled CLI, and the server reuses the cookies.
atlassian_auth_status tells an agent which service needs a login before it fails mid-task.ATLASSIAN_SEED_FROM_CHROME_PROFILE), so the first login is often one click.atlassian-cli) for scripts and agents: get and search issues and pages.ATLASSIAN_BROWSER_AUTH_ENABLED=false.From PyPI, with uv installed:
JIRA_URL=https://jira.example.com uvx --from atlassian-browser-mcp atlassian-cli login jira # once per service, opens your browser
CONFLUENCE_URL=https://confluence.example.com uvx --from atlassian-browser-mcp atlassian-cli login confluence JIRA_URL=https://jira.example.com CONFLUENCE_URL=https://confluence.example.com uvx atlassian-browser-mcp
From a checkout, with the exact versions in uv.lock:
git clone https://github.com/GeiserX/atlassian-browser-mcp.git && cd atlassian-browser-mcp
UV_PROJECT_ENVIRONMENT=.venv-atlassian-browser uv sync --frozen # installs the versions in uv.lock JIRA_URL=https://jira.example.com CONFLUENCE_URL=https://confluence.example.com ./atlassian-cli login jira # once per service
Needs Python 3.11+, uv, Google Chrome and a display for the login. The server never opens a browser window. It reads the saved cookies; when they are missing or expired it returns an AuthRequiredError asking you to run the login again, and on macOS it also looks for a live session in your installed browsers in the background, so a retry a few seconds later may already work. Seeding from Chrome and every setting are in Getting started.
Every client runs the server over stdio. Set only the URL of a service you use. The examples use the PyPI package through uvx; from a checkout, replace uvx atlassian-browser-mcp with /path/to/atlassian-browser-mcp/run-atlassian-browser-mcp.sh.
Claude Code
claude mcp add --transport stdio atlassian \
--env JIRA_URL=https://jira.example.com \ --env CONFLUENCE_URL=https://confluence.example.com \ -- uvx atlassian-browser-mcp
Claude Desktop (claude_desktop_config.json) and Cursor (~/.cursor/mcp.json or .cursor/mcp.json)
{
"mcpServers": { "atlassian": { "command": "uvx", "args": ["atlassian-browser-mcp"], "env": { "JIRA_URL": "https://jira.example.com", "CONFLUENCE_URL": "https://confluence.example.com" } } } }
VS Code (.vscode/mcp.json)
{
"servers": { "atlassian": { "type": "stdio", "command": "uvx", "args": ["atlassian-browser-mcp"], "env": { "JIRA_URL": "https://jira.example.com", "CONFLUENCE_URL": "https://confluence.example.com" } } } }
Codex (~/.codex/config.toml)
[mcp_servers.atlassian]
command = "uvx" args = ["atlassian-browser-mcp"] env = { JIRA_URL = "https://jira.example.com", CONFLUENCE_URL = "https://confluence.example.com" }
Protocol. Speaks MCP 2025-11-25, 2025-06-18, 2025-03-26 and 2024-11-05 over stdio, which covers clients that open with an initialize handshake (Claude Code, Claude Desktop, Cursor and VS Code today). MCP 2026-07-28 is not supported yet: upstream mcp-atlassian pins mcp<2 and fastmcp<4 (https://github.com/sooperset/mcp-atlassian/issues/1541), and this server follows as soon as upstream moves. It runs on stdio only and refuses to start with any other FASTMCP_TRANSPORT.
Write tools act as you, with your SSO identity. To hide all of them (create, update, delete, transition, upload, restrictions), set READ_ONLY_MODE=true in the server's env. Upstream enforces it when listing and when calling. ENABLED_TOOLS (a comma-separated allow-list of tool names) and TOOLSETS narrow the list further; see Configuration.
| Tool | What it does | Hints |
|---|---|---|
atlassian_auth_status | Whether each service's saved session exists and works: jar age, cookie expiry, a live check, the next step | read-only |
atlassian_login | The exact atlassian-cli login command to show the user | read-only |
jira_* (63) | Every upstream Jira tool: issues, search, comments, transitions, sprints, boards, versions, links, worklogs, attachments, service desk, forms | 38 read-only, 25 write |
confluence_* (35) | Every upstream Confluence tool: pages, search, comments, labels, attachments, history, templates, restrictions | 20 read-only, 15 write |
Every tool carries a title and all four annotation hints (read-only, destructive, idempotent, open-world), so clients can auto-approve reads and confirm writes. The full list and the CLI are in Usage.
GPL-3.0-or-later