Power your agent
คำสั่งของผู้เผยแพร่เอง ไม่ใช่เอนด์พอยต์ของ PlugMCP
Host a copy with our keys and logs.
npx -y [email protected]Claude Code ไม่มีลิงก์ติดตั้งสำหรับคำสั่งแบบโลคัล ให้คัดลอกคำสั่งด้านล่างแล้วรันในเทอร์มินัล
claude mcp add layero -- npx -y layero@0.11.27รันในเทอร์มินัล
{ "mcpServers": { "layero": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
URL แบบรีโมตหรือคำสั่งบนเครื่องของคุณ? การติดตั้งสองแบบต่างกันอย่างไร
รันบน PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
เปิดจากผู้เผยแพร่
CLI for Layero — hosting with build servers in Russia. Deploy a directory or connect a repository with one command; every command speaks JSON for AI agents and CI.
Source: github.com/LayeroInfra/cli (the CLI's home; mirrored to gitverse.ru/layero/cli). Docs: .
Layero CLI — deploy a directory or a repo to hosting with build servers in Russia. JSON events and exit codes for AI agents and CI.
The canonical agent skill lives in
LayeroInfra/layero-agents:
npx skills add LayeroInfra/layero-agents, or for Claude Code
claude plugin marketplace add LayeroInfra/layero-agents && claude plugin install layero@layero.
Agent docs: docs.layero.ru/agents.
Three paths — pick by situation:
main
= production.
npx layero@latest projects create --repo github:owner/repo --json
npx layero@latest deploy --json
diagnose, logs, rollback,
envs list, domains, env, analytics, data — or the MCP server
https://mcp.layero.ru/mcp (npx -y add-mcp https://mcp.layero.ru/mcp).layero deploy --json prints {"event":"ready","url":"https://…"} on
success. Show url to the user as-is. Not logged in? The command starts the
browser device flow itself and prints auth_required with a url — render
it as a link and keep waiting. No account at all? --claim (see below).
npx layero@latest deploy # one-shot, always the current version
npm install -D layero # project-local
Requires Node.js ≥ 20. Do not npm install -g layero: without @latest a
bare npx layero call then runs the globally installed copy for years.
layero login # device flow: prints a URL + code, sign in once (email code or Yandex ID)
cd my-site
layero deploy --dry-run # how the platform will build this folder; uploads nothing, no login needed
layero deploy # packs, uploads, the platform builds and ships
The first layero deploy in a directory creates a project and links it via
./.layero/project.json. Later runs reuse the same project.
A plain
layero deployis not a preview. For a project created from the CLI, direct uploads auto-promote: every run replaces what visitors see atready.url.--branchis refused withbranch_unsupported(exit 4): archive uploads always land in the reservedclienvironment, so the flag cannot give you a preview. Isolated previews come from pushing a branch of a connected repository —layero projects create --repo ….
Run layero <cmd> --help for full options.
layero deploy flags--dry-run — print how the platform will build this folder and exit:
framework, build command, output folder, where each value comes from
(layero.json, project settings, package.json, a framework default), and
a hint / next_action when the folder is a monorepo, a frontend +
backend pair, a custom build script or a server. Uploads nothing, needs no
login (with a login it also reads the settings of a linked project).--type <preset> — framework override: vite, vitepress, next,
astro, cra, sveltekit, nuxt, gatsby, docusaurus, eleventy
(alias 11ty), hugo, static, generic; runtime kinds node_web,
python_web, flask, streamlit, gradio, ssr_next (aliases
express, fastapi, django, …). Optional — detected by the
platform when omitted. static serves the files as they are and never runs
a build; generic runs your own build command (layero.json
buildCommand or the build script) and serves the folder with
index.html.--prebuilt [dir] — ship an already-built artifact instead of building
remotely. Without an argument, picks the first existing of dist/,
build/, public/, out/, _site/, .output/public/,
docs/.vitepress/dist/, .vitepress/dist/.--root <dir> — monorepo: the app lives in a subdirectory (--root apps/web).--name <name> — project name (only on first deploy).--project <id_or_slug> — deploy into an existing project, ignoring
./.layero/project.json (use this in CI, not --name).--prod — only for a project with a connected repository: publish this
upload at the live address (without it the upload lands in the project's
separate cli environment). A project without a repository is always
published live.--branch <name> — refused (branch_unsupported, exit 4), see above.--claim — deploy without an account: a temporary site for 1 hour
plus a claim_url for a human to take it over. Static sites and SPAs only:
a server app (SSR, fullstack, container) is refused before anything is
uploaded (claim_static_only, exit 4). Only with this flag: without a
token and without --claim, deploy asks for a login (auth_required),
in an agent environment too. With --project it is refused
(claim_with_project, exit 4). In CI a missing LAYERO_TOKEN stays an error.--org <slug> — organization for first-time project creation.--yes / -y — non-interactive mode.--json — JSON-lines events on stdout (for agents and CI).--claim)npx layero@latest deploy --claim --json
The platform creates a temporary project and a token for it; the CLI deploys
with that token and prints, before ready:
{"event":"claimable","url":"https://k3v9q2m8x7w1c4r6t0y5u2ze.layero.app","claim_url":"https://app.layero.ru/claim?code=…","expires_at":"…"}
The site lives for 1 hour, then stops answering and is deleted. Only static
sites and SPAs are accepted: a server app needs an account. The address is
random (the folder name and --name do not go into it), and the site is
closed to search engines: robots.txt with Disallow: / and
X-Robots-Tag: noindex, nofollow. A human opens claim_url, signs in and
takes the project into their account — the CLI cannot accept a claim by
itself.
The temporary token, the claim code and the link are saved in
~/.layero/config.json (readable by you only); .layero/project.json gets
only the folder's link to the project, because that file goes to git and a
claim code there would let anyone take the site. layero deploy in the same
directory keeps updating the same site until the claim expires or is
accepted. Claiming the site clears what was set up without an account —
environment variables, deploy hooks and build settings.
layero claim status shows where things stand; layero claim accept opens
the page in a browser (in agent mode it prints the link).
layero deploy --dry-run (and layero init, and the detected event of every
deploy) read your project on disk with the same rules the platform uses. This
is advice, not a decision: the platform detects again on the uploaded
files, and nothing the CLI guesses is saved to the project — only what you name
(--type, --root, fields you write into .layero/project.json or
layero.json). confident: false means the folder was not recognised; read
hint and next_action (an app in a subfolder → --root <dir>, frontend +
backend → layero.json with both halves, a custom build script →
"framework": "generic").
layero.json — pin the settings in the repositoryAuto-detection above is a default, not a decision. Drop a layero.json at the
root of the repository and Layero uses what you set there instead — for the
CLI, the dashboard and pushes alike. It beats any dashboard setting.
{
"$schema": "https://layero.ru/schema/layero-v2.json",
"framework": "vite",
"installCommand": "npm ci",
"buildCommand": "npm run build",
"outputDirectory": "dist",
"nodeVersion": "22"
}
Every field is optional; {} is valid. An error in the file never fails a
build: unreadable values become warnings in the build log.
Full reference: https://docs.layero.ru/deploys/layero-json
sources and projects createlayero sources list # providers + connections
echo "$GITVERSE_TOKEN" | layero sources connect gitverse --token-stdin
layero sources repos <connection_id>
layero projects create --repo gitverse:acme/site --branch main --json
GitHub is connected by installing the Layero GitHub App in the dashboard; the
other providers take a personal access token. projects create validates the
repository against the connection, creates the project and installs the
webhook. If the provider refuses the webhook (token permissions, or
SourceCraft, which has no outgoing webhooks), the CLI says so with
webhook_unavailable and the URL to register by hand — the repository is
connected either way, only push-triggered builds wait for the webhook.
Once linked, the command finishes the setup wizard itself: it takes the
detection hint (framework, build command, output directory), applies it and
starts the first build — exactly what the dashboard's "Start deploy" button
does (setup_applied, deploy_started). --no-deploy leaves the project in
the wizard (setup_pending with the dashboard URL). If detection or setup
fails, the project is still created and the exit code is 0: setup_failed
carries the reason and the wizard URL.
layero hooks create strapi-content # preview-target, default branch
layero hooks create publish --prod # production-target hook
layero hooks create staging --branch=dev # explicit branch
layero hooks list
layero hooks delete <id> # revoke immediately
The URL (https://api.layero.ru/hooks/<token>) is a credential: anyone who
has it can start a build. Rotate with delete + create.
--prebuilt)layero deploy --prebuilt # auto-pick the output directory
layero deploy --prebuilt ./dist # or point at a specific one
Only the files inside that directory are uploaded and shipped verbatim — no
detect, no install, no build. .gitignore is not applied on this path
(.env*, .git, node_modules and the rule files are still excluded), so
name the directory explicitly rather than using ..
layero login opens a browser — there isn't one on a runner, so a pipeline
authenticates with a long-lived token. Create it with layero token create ci
or at app.layero.ru/settings/cli:
LAYERO_TOKEN=... npx layero@latest deploy --project <slug> --json --yes
LAYERO_TOKEN is read before ~/.layero/config.json. --yes skips the
confirmation that would otherwise wait forever. --project, not --name:
--name only names a project on creation, and a clean checkout without
.layero/project.json would create a new project on every run.
On GitHub Actions there is an official action:
- uses: LayeroInfra/deploy-action@v1
with:
token: ${{ secrets.LAYERO_TOKEN }}
prod: true
If the repository is already connected to a Layero project, a push builds it
automatically — reach for CI only when the build itself needs secrets the
platform does not have, then ship the result with --prebuilt.
Full guide: https://docs.layero.ru/cli/github-actions
layero switches to non-interactive, structured output when any of these holds:
--json flag or LAYERO_JSON=1CURSOR_AGENT, CLAUDECODE, LAYERO_AGENT env varsCI=1 (non-interactive only; JSON-lines requires explicit opt-in)Every command emits events — one JSON object per line on stdout:
{"event":"auth_required","url":"…","user_code":"…"}
{"event":"authorized","user":"…"}
{"event":"me","id":"…","username":"…","email":"…"}
{"event":"projects","projects":[{"slug":"…","url":"https://…","repo":null,…}]}
{"event":"organizations","organizations":[{"slug":"…","kind":"personal","role":"admin"}]}
{"event":"project_created","project_id":"…","slug":"…","organization":"…","url":"…","repo":"…","branch":"…"}
{"event":"project_linked","project_id":"…","slug":"…","url":"…"}
{"event":"source_connected","org":"…","connection_id":"…","provider":"…","account":"…"}
{"event":"webhook_installed","project":"…"} | {"event":"webhook_unavailable","project":"…","url":"…","hint":"…"}
{"event":"setup_applied","project":"…","framework":"…","build_cmd":"…","output_dir":"…","layero_found":false}
{"event":"setup_pending","project":"…","url":"https://app.layero.ru/projects/…/setup","hint":"…"} | {"event":"setup_failed","project":"…","reason":"…","url":"…","hint":"…"}
{"event":"sources","org":"…","providers":[…],"connections":[…]}
{"event":"source_repos","org":"…","connection_id":"…","repos":[…]}
{"event":"environments","project":"…","environments":[{"branch":"main","url":"https://…","production":true,…}]}
{"event":"detected","framework":"…","build_cmd":"…"|null,"output_dir":"…"|null,"confident":true,"sources":{…},"hint":"…","next_action":"…"}
{"event":"plan","framework":"…","build_cmd":"…"|null,"output_dir":"…"|null,"sources":{…},"creates_project":true,"replaces_live_site":true,…} (--dry-run)
{"event":"packing","files":N,"bytes":N,"sha256":"…"}
{"event":"uploading"} {"event":"uploaded","archive_key":"…"}
{"event":"deploy_started","deploy_id":"…"}
{"event":"build_log","line":"…","stream":"…"}
{"event":"stage","name":"…"}
{"event":"claimable","project_id":"…","slug":"…","url":"…","claim_url":"…","expires_at":"…"}
{"event":"ready","url":"…","dashboard_url":"…","deploy_id":"…"}
{"event":"hooks","project":"…","hooks":[…]} {"event":"hook_created",…} {"event":"hook_deleted",…}
{"event":"project_deleted","project_id":"…","slug":"…"}
{"event":"claim_status","code":"…","status":"…","claimed":false,"expires_at":"…","url":"…","claim_url":"…"}
{"event":"claim_accept","code":"…","claim_url":"…","opened":false}
{"event":"logged_out","config_path":"…"}
{"event":"error","code":"…","next_action":"…","message":"…"}
On ready, url is the live public site — show it as-is and never
rebuild the hostname from a template. ready comes once the address answers
with the site itself (edge_ready: true); edge_ready: false with screen
means the platform's own page still answered after the wait — the app did not
come up, read layero logs --runtime. dashboard_url is the management page,
not the site. preview_url and edge_eta_seconds are legacy fields.
Errors carry a stable code (auth_required, auth_expired, auth_timeout,
project_unknown, project_not_found, cli_deploys_disabled, invalid_type,
prebuilt_no_dir, prebuilt_no_index, branch_unsupported,
deploy_not_started, deploy_failed, internal, and command-specific ones)
plus a next_action hint. The failure code is assembled as deploy_<status>
and a deploy only has ready, building, failed and cancelled — so
deploy_error, deploy_timed_out and not_logged_in do not exist.
Full reference: https://docs.layero.ru/cli/json-events
layero deploy honours .gitignore and .layeroignore. Always excluded:
node_modules, .git, dist, build, .next, .env*, .DS_Store, and
the rule files themselves. Maximum archive size is 500 MB.
~/.layero/config.json (chmod 600). Claim tokens of temporary
projects live there too, keyed by project../.layero/project.json — project_id, slug,
organization_slug, apex_hostname, claim are managed by the CLI;
framework_hint, build_cmd, output_dir, analytics_enabled,
env_vars are yours: the CLI never writes guesses there, and what you put
there is applied to a new project as your choice.This repository is the CLI's home (since 2026-09-18; before that it was the
cli/ directory of the platform monorepo and this repo was a read-only
mirror). Issues and pull requests are welcome here. make check runs the
build, the unit tests and the consistency gates; npm test alone runs the
tests, npm run build compiles with tsc. Releases are published to npm by
.github/workflows/publish.yml on a vX.Y.Z tag. Contributor rules for
humans and agents: AGENTS.md; how the package is put together: ARCH.md.
Layero CLI — deploy a directory or a repo to hosting with build servers in Russia. JSON events and exit codes for AI agents and CI.
{ "mcpServers": { "layero": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}เครื่องมือ
ยังไม่มีรายการเครื่องมือ เซิร์ฟเวอร์แบบรีโมตจะได้รายการหลังการตรวจสอบสถานะครั้งแรก
FAQ
คัดลอกคอนฟิก stdio สำหรับ Cursor, Claude Desktop, Claude Code หรือ VS Code ได้ในหน้านี้ คอนฟิกนี้ใช้การติดตั้งของผู้เผยแพร่ ไม่ใช่พร็อกซีของ PlugMCP
ระเบียนนี้ไม่ได้ระบุตัวแปรสภาพแวดล้อมที่จำเป็น
Layero ทำงานแบบโลคัลผ่าน stdio
ตรวจสอบล่าสุดเมื่อ 2026-11-14 Layero CLI — deploy a directory or a repo to hosting with build servers in Russia. JSON events and exit codes for AI agents and CI.