WebCake Landing MCP
English Β· TiαΊΏng Viα»t

Describe a landing page in plain words β your AI builds it, checks it, and ships it straight to WebCake.
β If this saves you an afternoon of dragging boxes, give it a star β it's a one-dev project and every star keeps it alive.
"Build a landing page for my coffee shop β a hero with a sign-up button, a 3-feature section, and a lead form. Save it to my workspace."
β¦and a real, editable WebCake page appears in your account. No dragging boxes, no learning the schema, no hand-writing JSON.
π§© How it works
This server is the bridge between your AI assistant and WebCake. The AI never guesses what a WebCake
page looks like β it asks this MCP, which knows the entire element model, validates the result, and saves it for you.
You AI assistant webcake-landing MCP WebCake
ββββββββ prompt ββββββββββββββ tools ββββββββββββββββββββββββ API ββββββββββββ
β idea β ββββββββΊ β Claude / β βββββββΊ β β’ knows the element β βββββΊ β a real β
β β β Cursor / β β model + AI hints β β editable β
β β ββββββββ β Windsurf β βββββββ β β’ builds + validates β βββββ β page in β
ββββββββ page URL ββββββββββββββ result β β’ saves to your acct β β WebCake β
ββββββββββββββββββββββββ ββββββββββββ
- You ask in plain language β goal, brand, sections, CTA, form fields.
- The AI learns the model from the MCP: the element catalog, the absolute-positioning canvas, the event vocabulary β so it builds a real WebCake page, not a guess.
- It assembles + validates the full
{ page, popup, settings, options } JSON. validate_page catches off-canvas boxes, dangling CTAs, and missing form fields before anything is saved.
- It persists to your WebCake account β dry-run preview first, then for real.
- You get an editor link β open it, tweak, publish. The AI did the heavy lifting.
Why it's reliable
| |
|---|
| π Knows the real model | Serves WebCake's actual element catalog (40+ types β hero, form, countdown, gallery, product listβ¦), each with its exact specials and AI hints, drawn straight from the editor's renderers. |
| β
Validates before saving | Structural + semantic checks (unique ids, on-canvas layout, working CTAs, unique form fields) so the page isn't broken when it lands. |
| π‘οΈ Safe by default | Every write is dry-run first (preview the request, token masked) β nothing touches your account until you confirm. |
| βοΈ Edits surgically | Ask for one change ("make the CTA green") and it edits only that element β every other id, coordinate, and block stays exactly as it was. |
π‘ Lead-gen, events, invitations, app promos β or selling COD/online? It speaks WebCake's commerce model too (product lists, variations, cart).
β¨ What you can build
One sentence to your AI β a finished, editable WebCake page. A taste of what people ship with it:
| Just say⦠|
|---|
| π§² Lead-gen landing | "A SaaS waitlist page β hero, 3 benefits, an email-capture form." |
| π COD / online store | "A one-product page for my skincare serum β gallery, price, variations, an order form with cart." |
| ποΈ Event / webinar | "A registration page for Saturday's webinar β countdown, agenda, sign-up form." |
| π Invitation | "A wedding invite β names, date, a map, an RSVP form." |
| π± App promo | "A page for my fitness app β phone mockups, feature list, App Store + Google Play buttons." |
| β‘ Flash sale | "A flash-sale page β big countdown, discounted product grid, a sticky Buy button." |
| π Link-in-bio | "A link-in-bio for my creator profile β avatar, short bio, 5 link buttons, socials." |
| π Product launch | "A launch page for v2 β hero, what's-new list, an early-access form." |
β¦then "make the CTA green" or "add a 4th feature" and it edits only that block β every other id and coordinate stays exactly where it was.
π€ Works in Claude Desktop, Claude Code, Cursor, Windsurf, Augment, Codex, Antigravity, Gemini CLI, Cline, Kiro, OpenCode, or any MCP-capable client β and the reference + generation tools need zero setup, so you can try it before ever pasting a token.
Under the hood
MCP (Model Context Protocol) server that teaches AI agents how to build a complete
WebCake landing-page source JSON from a requirement β and persist it to a WebCake backend.
It exposes the element catalog, per-element usage hints + specials, the full page JSON Schema,
valid element/page skeletons, a page validator, and tools to create or edit pages on the backend.
The AI agent produces the full { page, popup, settings, options, cartConfigs } JSON; create_page
persists it and auto-publishes (build + publish_html) so the preview renders immediately (the edit
tools save source-only β re-publish via publish_page after edits).
| Method | Best for | Auth |
|---|
| npx (local) β runs on your machine | Personal daily use, full control | browser login, a JWT, or none (reference tools) |
| Hosted URL β use our live server, nothing to install | No Node.js, teams, the claude.ai dialog | your personal ?jwt= link / x-webcake-jwt header |
The reference + generation tools (get_generation_guide, list_elements, validate_page, β¦) and the ingest tools (ingest_html, ingest_url β turn an existing HTML or URL into a layout anchor so the AI can recreate or adapt it) work with zero config; only the persistence tools (create_page, update_page, add_section, patch_page, publish_page, list_pages, find_pages, get_page, list_organizations) need a token. Credentials resolve in order: per-request header β env var β saved auth.json (login).
π Get connected β the 2 main ways
Pick one. Both hand your AI tool (Claude, Cursor, β¦) the full Webcake landing toolkit. No coding.
β npx β runs on your machine (recommended for personal use)
Zero install, always the latest version, needs Node.js 18+. One line grabs your token and writes the IDE config:
npx -y webcake-landing-mcp install
npx -y webcake-landing-mcp install --dev
npx -y webcake-landing-mcp install --ide all --jwt <your-jwt>
npx -y webcake-landing-mcp uninstall
install asks nothing. It connects to production, opens your browser so you can click Connect, then adds the server
to every AI app it finds on this computer. Your other servers are kept, and any config file it changes is backed up next to
the original as <file>.webcake-landing-backup. No AI app found? It says so and exits with an error: install one and run it again,
or name it with --ide.
Need staging or local? --dev opens the step-by-step wizard, or skip it with --env staging / --env local.
Installed with an older version? Run install again. Older versions wrote VS Code's config under the wrong key, so the
server never appeared there, and registered Claude Code for the current folder only.
It writes a webcake-landing entry into the right config file for each target: claude-desktop,
claude-code, cursor, windsurf, augment (VS Code), codex, antigravity, gemini (Gemini CLI),
cline, kiro, opencode, or all. Flags: --dev, --ide, --env, --jwt, --org-id,
--api-base/--app-base, --npx/--local, -y β see install --help.
Just want to run the server (configure by hand later)? npx -y webcake-landing-mcp
π οΈ Hand-written per-IDE config, shell-script installers (install.sh/install.ps1), or a cloned
local build β docs/manual-install.md.
β‘ Remote URL β¦/mcp?jwt= β hosted, nothing to install
The server is already live at https://mcp.toolvn.io.vn/mcp β no Node.js, no machine to keep awake.
Grab your personal link (your token is baked in) and paste it into your client's Add custom connector / config:
https://mcp.toolvn.io.vn/mcp?jwt=<YOUR_TOKEN>
Two ways to get the link:
Optional extras: &env=prod, &org_id=β¦, &api_base=β¦. Hand each teammate a link with their own jwt β
per-user, no OAuth. Clients that support headers should send the token as x-webcake-jwt instead of
putting it in the URL β the full header β env mapping is in docs/configuration.md.
β
Best for: no Node.js, team/shared use, the claude.ai connector dialog (URL-only, no headers).
β οΈ The link contains your personal token β treat it like a password, always use HTTPS.
βοΈ Configuration
The quick version β only the persistence tools need any of this:
npx -y webcake-landing-mcp login
β¦or set WEBCAKE_ENV (local | staging | prod β fills in all base URLs) + WEBCAKE_JWT.
For publish_page to actually put a page live, a build host is needed (it renders the
app/app_css that the live publish_html route requires):
prod preset auto-configures https://build.webcake.io β no extra setup (the preset applies when the env resolves to prod: WEBCAKE_ENV=prod, --env prod, or x-webcake-env: prod).
- For staging/local, set
WEBCAKE_BUILD_BASE=<url> or send the x-webcake-build-base header per request.
- Without it,
publish_page falls back to a legacy source-only save with rendered:false, live:false + a warning β nothing goes live.
- A page is only permanently live with a
custom_domain; without one the returned /preview/<page_id> link expires ~10 minutes after the publish.
Everything else β the full env-var table, environment presets, per-request headers for the hosted
server, the login browser flow (+ backend contract), and how to grab a JWT by hand β lives in
docs/configuration.md.
π Docs
23 tools in five groups β full descriptions in docs/tools.md:
| Group | Tools | Needs |
|---|
| Reference | get_generation_guide Β· list_elements Β· get_element Β· get_page_schema | nothing |
| Generation | new_element Β· new_page_skeleton Β· validate_page Β· layout (exact centering/row/grid/stack coordinates, both breakpoints) | nothing |
| Media | search_images (real Pexels stock photos) Β· get_icon_svg (Material Symbols / Font Awesome icon names β inline SVG via Iconify) Β· upload_images (re-host external images, data: URIs, or local file paths from the user's machine β files into your media collection when Webcake creds and an org are available; the org is required and must match the page's, so settle it up front β 1 org auto-selects, 2+ with none chosen returns organization_required + the list; no JWT β the public CDN) Β· render_preview (screenshot a page/URL so the model can see + compare it) | nothing |
| Ingest | ingest_html Β· ingest_url (recreate an existing page) | nothing |
| Persistence | list_organizations Β· create_page Β· list_pages Β· find_pages Β· get_page Β· update_page Β· add_section Β· patch_page Β· publish_page | WEBCAKE_API_BASE + WEBCAKE_JWT |
Every write defaults to dry_run=true β it previews the exact request (token masked) and only
touches your account when you re-run with dry_run=false.
Suggested prompt
Build me a WebCake landing page for <brand/offer>. Use the webcake-landing MCP:
call get_generation_guide, new_page_skeleton, then get_element for each element type you use,
assemble the { page, popup, settings, options } JSON, validate_page until zero errors,
then create_page (dry-run first).
β Like the idea? Drop a star
This is a solo, open-source project β every β genuinely keeps it moving and helps other builders discover it.
- β Star the repo β 2 seconds, huge motivation.
- π Open an issue β a bug, a missing element type, or just an idea.
- π Share it with anyone still building landing pages box by box.

Built with β€οΈ for the WebCake community. Thanks for being here.