Browser Use
Browser agents and stealth cloud browsers for AI: run web tasks, fill forms, extract data.
제작자가 쓴 설명(영어 원문)
Publish HTML to private, unguessable, replace-in-place URLs from any MCP-compatible AI agent.
이 세 가지는 절대 하나의 점수로 합치지 않아요. 의견, 측정, 인기는 서로 다른 것이니까요.
이 서버는 아직 쉬운 말로 된 요약을 쓰지 못했어요. 아래는 레지스트리에서 읽어 온 사실이에요.
쓰고 있는 AI 앱을 고른 다음 설정을 복사하세요.
필요한 것
어떤 앱을 쓰세요?
위에서 앱을 고르세요. 기억해 두었다가 다음부터는 버튼에 “…에 설치”라고 표시해요.
웹, 데스크톱, 모바일의 Claude에서 쓸 수 있어요:
https://api.stacktr.ee/mcpClaude Desktop → Settings → Developer → Edit Config를 열고 claude_desktop_config.json에 이 내용을 붙여 넣으세요
{
"mcpServers": {
"publish": {
"command": "npx",
"args": [
"-y",
"stacktree-mcp"
]
}
}
}터미널에서 이 명령어를 실행하세요
claude mcp add --transport http publish https://api.stacktr.ee/mcpclaude mcp add publish -- npx -y stacktree-mcp프로젝트의 .cursor/mcp.json 또는 ~/.cursor/mcp.json에 붙여 넣으세요
{
"mcpServers": {
"publish": {
"url": "https://api.stacktr.ee/mcp"
}
}
}프로젝트의 .cursor/mcp.json 또는 ~/.cursor/mcp.json에 붙여 넣으세요
{
"mcpServers": {
"publish": {
"command": "npx",
"args": [
"-y",
"stacktree-mcp"
]
}
}
}브라우저가 앱을 열지 물어본 다음, 앱이 설정을 보여 주고 확인을 요청해요. 키 값은 비워 두니 직접 입력하세요.
프로젝트의 .vscode/mcp.json에 붙여 넣으세요
{
"servers": {
"publish": {
"type": "http",
"url": "https://api.stacktr.ee/mcp"
}
}
}프로젝트의 .vscode/mcp.json에 붙여 넣으세요
{
"servers": {
"publish": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"stacktree-mcp"
]
}
}
}브라우저가 앱을 열지 물어본 다음, 앱이 설정을 보여 주고 확인을 요청해요. 키 값은 비워 두니 직접 입력하세요.
ChatGPT는 인터넷에 있는 서버에만 연결할 수 있어요:
https://api.stacktr.ee/mcp메뉴 이름은 요금제나 시기에 따라 조금 다를 수 있어요.
선택한 앱은 이 기기의 쿠키에만 저장돼요. 계정은 필요 없어요.
이 설정은 제작자의 레지스트리 정보로 만들었어요. README와 다르면 README를 따르세요.
서버가 도구 목록을 보여 주기 전에 로그인을 요구해요.
누구나 볼 수 있는 정보를 점검했고, 코드는 읽지 않았어요. 해당하지 않는 항목은 개수에서 뺐어요. · 2026년 11월 8일 점검
레지스트리 이름이 도메인 stacktr.ee 기준으로 확인되었어요(레지스트리가 도메인 소유를 확인해요)
아직 GitHub에서 이 저장소를 읽어 오지 않았어요
아직 GitHub에서 이 저장소를 읽어 오지 않았어요
43ms 만에 응답(HTTP 401)
OAuth 로그인을 지원한다고 밝혀서, AI 앱이 로그인 화면으로 안내해 줘요
아직 점검하지 않았어요. 엔드포인트를 점검할 때 도구 목록을 읽어요
레지스트리에서 이 카테고리의 활성 서버 전체와 비교했어요. 스타는 게시자가 저장소를 소유한 경우에만 세요.
해당하는 기본 점검 3개 중 3개를 통과했어요. 코드는 읽지 않았어요. 키나 토큰을 넣기 전에 서버가 어떤 권한을 요구하는지 확인하고, 꼭 필요한 권한만 있는 키를 쓰세요.
레지스트리에 필수 키가 나와 있지 않아요. 다만 처음 쓸 때 로그인을 요구하는 서버도 있어요.
둘 다 가능해요. 내 컴퓨터에 설치해도 되고, 클라우드 버전에 연결해도 돼요.
GitHub 저장소는 마지막으로 – 변경됐고, 최신 버전(0.8.3)은 17일 전 레지스트리에 올라왔어요.
Claude Desktop, Claude Code, Cursor, VS Code처럼 MCP를 지원하는 앱에서 쓸 수 있어요. URL로 연결되기 때문에, 원격 MCP 서버를 지원하는 앱(예: 개발자 모드의 ChatGPT)에서도 쓸 수 있어요.
같은 카테고리에는: Browser Use, Chrome DevTools MCP, MCP Server Browser
같은 카테고리에서 다른 제작자가 만든 서버예요. 스타 순으로 정렬했어요.
카드 보는 법:★ GitHub 스타i세 기둥: 리뷰 · 점검 · 사용 정도i업데이트 = 최신 버전을 레지스트리에 공개한 날i
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
제작자의 저장소에서 쓰인 그대로 가져왔어요. PopNx MCP가 검토하거나 보증한 내용이 아니에요. HTML, 이미지, 스크립트는 제거했어요. 링크를 누르면 이 사이트를 벗어나요.
MCP server for stacktr.ee. Publish HTML artifacts from any MCP client (Claude Desktop, Claude Code, Cursor, Continue, etc).
If you have a coding agent open, hand it this line and it does the rest — installs, verifies the connection, and learns the tool surface:
Fetch and follow the setup instructions at https://stacktr.ee/prompt.md
Works in any agent that can fetch a URL. The instructions are plain Markdown; read them first if you like.
Add to your MCP client config:
{
"mcpServers": { "stacktree": { "command": "npx", "args": ["-y", "stacktree-mcp"], "env": { "STACKTREE_API_KEY": "stk_live_..." } } } }
Generate an API key at https://app.stacktr.ee.
This is the stdio bridge, for clients that cannot speak streamable HTTP. If yours can, connect straight to the hosted server at https://api.stacktr.ee/mcp, which exposes the same 42 tools and takes the same key:
curl -sS -X POST https://api.stacktr.ee/mcp \
-H "Authorization: Bearer $STACKTREE_API_KEY" \ -H 'accept: application/json, text/event-stream' \ -H 'content-type: application/json' \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'
No browser and no OAuth flow is required to do that: an stk_live_ key is a valid credential on the MCP endpoint, exactly as it is on the REST API. OAuth 2.1 with Dynamic Client Registration is also accepted, for connectors acting on behalf of a signed-in human; session cookies are refused. Send one credential.
No key, and no human to ask? POST https://api.stacktr.ee/provision buys one over x402. A human nearby but no browser? POST https://api.stacktr.ee/api-keys/device-code, print the returned verification_url_complete for them, and poll /api-keys/device-code/poll until it returns the key (RFC 8628).
| Tool | What it does |
|---|---|
publish_html | Publish HTML; returns { url, id, expires_at, ... }. Pass client to file it under a client space. |
update_site | Replace the HTML of an existing site in place; URL is preserved. |
get_site | Read a site's current HTML source — edit it, then update_site. |
set_password | Add or clear a passcode gate. Works on every plan. |
set_expiry | Set hours-from-now expiry, or null for never. Clamped to the plan ceiling. |
set_email_gate | Restrict viewers to an email domain (one-time magic-link verify). Paid plans only. |
set_client_feedback | Let the people you send a page to comment on it (words, images, video) and react, with no account. |
set_agentation | Older developer mark-up toolbar. Prefer set_client_feedback. |
list_sites | List sites owned by this API key. Paged (has_more + cursor); filter with client. |
delete_site | Take a page down. The link dies at once; the content is kept 30 days, then destroyed. |
restore_site | Put a deleted or expired page back at the same URL, inside those 30 days. |
claim_site | Adopt a page published without an account, using its claim_token. Same URL; a claim counts as a publish. |
get_content | Read a page back as html (exact stored source, editable) or text (stripped, cheap to read). |
get_me | This key's account, plan, counts and enforced limits. Read limits from here, never from a description. |
link_wallet | Link your wallet so pages you publish are owned by your account. |
list_feedback | Read client comments, unresolved first, each with what it is on and whether it is still on the page. |
create_share_link | Mint a link addressed to one person — every open through it is attributed to that name. |
list_share_links | The links on a page, with attributed opens and when each was last opened. |
revoke_share_link | Kill one link; the page and every other link keep working. |
resolve_feedback | Mark a feedback item addressed, with an optional note. |
set_client | File a page under a client space by name or slug (auto-created), or null to detach. |
list_client_spaces | The client spaces on the account: page counts, bound hostname, portal state. |
create_client_space | Create a space up front. Rarely needed — publishing with client creates one. |
update_client_space | Rename, archive, or set the space-wide viewer gate every page under it inherits. |
delete_client_space | Remove a space. Its pages detach and keep their URLs. |
get_design_guide | The house design guide — read it before generating or restyling a page. |
get_passcode | Show the passcode on a page you own, to send it to the client again (passcodes set since 13 September 2026). |
list_versions | Earlier versions of a page, newest first: one snapshot per update. |
restore_version | Put an earlier version back at the same URL after a bad update. |
whos_it_for | Record who a page is for and get their own link. Sends nothing. |
unopened_pages | Pages, and named people, that have not opened what was sent an hour on. |
get_revision_brief | Everything that came back on a page, as the brief for its next version. |
publish_revision | Replace a page in place and resolve the comments it answers, in one step. |
resolve_comments | Resolve several comments at once, each with a note the reviewer sees. |
request_approval | Ask named people to approve a page; one private link each. |
get_approval_status | Who has approved the version that is live now, section by section. |
list_responses | What readers sent through the forms marked on a page. |
resolve_addressed_notes | Close the reviewers' notes an update answered, with one line each. |
open_html_file | Opens an .html file in ChatGPT desktop's file viewer. Here it only echoes the file back. |
stacktree_pages | Your 25 newest pages: ChatGPT's pages panel. Same data as list_sites. |
publish_card_status | Refreshes the publish card in ChatGPT and Claude. Here it returns the page's detail. |
publish_card_open | The publish card's Open button: a 15-minute owner preview link to a page. |
If your agent pays for publishes from its own wallet (x402 / MPP), link that wallet to your Stacktree account so every page it publishes — past and future — is owned there:
link_wallet({ code }) → returns the exact message to sign.personal_sign).link_wallet({ code, wallet, signature }) → links the wallet and adopts the pages it already published.Turn comments on with set_client_feedback and whoever you sent the link to can select words, or click an image, video or section, and leave a comment only the owner sees. list_feedback returns them, unresolved first, each with a one-line target and on_page (still in the page, or gone since an edit). Fix the page in place with update_site (same URL): its response lists which open comments no longer match the new version. Then resolve_feedback each answered one with a short note, which the client sees next to their comment.
Full policy: https://stacktr.ee/privacy. In short:
https://api.stacktr.ee, authenticated with your Stacktree API key. It collects nothing from your machine beyond that and keeps no local data.Every site gets an unguessable https://stacktr.ee/p/{token}/ URL. Pass public_slug to opt into https://{slug}.stacktr.ee/.
pii_check defaults to block — uploads matching common secrets and PII shapes (emails, SSNs, credit cards, OpenAI/GitHub/Stripe API-key prefixes) are refused. Pass warn to publish anyway (matches are flagged in the response), or off to skip the scan.
A key on the free plan publishes 3 pages in total, and each one expires 7 days after it is published. The count is lifetime, not concurrent: deleting a page or letting it expire does not give the slot back.
expires_in_hours: "never" is refused on a plan that caps page lifetime, not quietly shortened: 409 expiry_clamped, nothing published, and the body carries the date the page would have got. Pass accept_clamp: true to take the ceiling, or tell the user the plan cannot make the link permanent. A *number* longer than the ceiling is shortened rather than refused, with expiry_clamped: true in the response. Either way, read expires_at_iso off the response and quote that, never the value you asked for.
Passcodes (set_password) work on every plan, free included. Email gates (set_email_gate) and viewer numbers are on paid plans. Hitting a ceiling returns HTTP 402 with a stable plan_* code in error:
| Code | Means |
|---|---|
plan_lifetime_limit_exceeded | All 3 free pages used. Deleting one does not help. |
plan_site_limit_exceeded | Active-page cap reached. |
plan_password_not_available | Passcodes are not on this plan (they are on Free; a plan can still be without them). |
plan_viewer_gate_not_available | Email gates are not on this plan. |
plan_domain_not_available | Custom domains are not on this plan. |
GET /me — or the get_me tool — returns the calling key's own limits object. Read caps from there rather than hard-coding them. Current plans and prices: https://stacktr.ee/pricing.md.