# JustWebsite MCP Manage your JustWebsite account (websites, pages, feeds, previews, media, deployments, analytics, billing) with the same permissions and rules as the dashboard. Endpoint and OAuth resource: https://app.justwebsite.eu/mcp Protected resource metadata: https://app.justwebsite.eu/.well-known/oauth-protected-resource/mcp Authorization server metadata: https://app.justwebsite.eu/.well-known/oauth-authorization-server Dynamic client registration: POST https://app.justwebsite.eu/mcp/register (returns the shared public client; nothing to configure by hand). Clients without dynamic registration: use public client_id justwebsite-mcp token_endpoint_auth_method: none (no client secret) Login, consent and tokens are handled by the identity provider (token issuer https://auth.justwebsite.eu/application/o/justwebsite-mcp/); access tokens are opaque to clients. Grant: authorization_code, PKCE S256; request openid mcp:read, plus mcp:write to make changes. Include resource=https://app.justwebsite.eu/mcp at authorization AND token exchange. Request offline_access only if refresh is desired; it is not a resource scope. Scopes do not imply each other or local permissions. Product administration is not available over MCP. Request mcp:write explicitly: some clients ask only for the scopes an earlier 401 challenge advertised, ignore their configured scope set, and silently end up read-only. A tool call without its required scopes gets HTTP 403 with WWW-Authenticate error="insufficient_scope" and the required scope list, and a JSON body naming them ({"error":"insufficient_scope","requiredScopes":["mcp:read","mcp:write"]}); re-authorize with those scopes (step-up). Callbacks: HTTPS, or HTTP with exact hostname localhost or 127.0.0.1; variable ports/paths/query allowed. No credentials, fragments, whitespace, controls, backslashes or ambiguous/encoded hosts; maximum 4096 characters. Always inspect the destination hostname and full callback on the hosted consent screen. Authenticate every request with Authorization: Bearer . Browser cookies and ID/refresh tokens do not work. Streamable HTTP, stateless JSON responses, no session IDs or standalone SSE/replay. Current protocol 2025-11-25; initialize compatibility also supports 2025-06-18 and 2025-03-26. ## Working with tools Edits change working pages; the public site changes only after deploy. Typical flow: list_websites → list_pages/get_page → create_page/update_page → get_page_preview_url → deploy. Lists are bounded (default 25, maximum 100): pass nextCursor back as cursor with the same other arguments; null means no more. Expected failures return a tool error with { code, message, fieldErrors?, details? }. Do not automatically retry non-idempotent tools; check current state first. update_page and set_website_logos replace all fields: read first and send everything you want to keep. Preview tools return secret tokens/links; do not publish them. Media: put a media item's componentValue into a component slot. Upload small files (≤512 KiB) with upload_media_inline (base64), larger ones (≤50 MiB) with request_media_upload, then PUT the raw bytes to the returned single-use URL within 10 minutes (e.g. curl -sS --fail-with-body -T ./photo.jpg ""). Allowed: JPEG, PNG, GIF, WebP, SVG, MP4, WebM. Billing tools return hosted provider URLs for the user to open in a browser; call refresh_billing afterwards. ## Tools - whoami [mcp:read]: Current identity - get_account [mcp:read]: Account, limits and usage - get_billing [mcp:read]: Billing state - refresh_billing [mcp:read mcp:write]: Refresh billing - begin_verification_checkout [mcp:read mcp:write]: Start verification checkout - begin_subscription_checkout [mcp:read mcp:write]: Start Premium S checkout - open_billing_portal [mcp:read mcp:write]: Open billing portal - list_websites [mcp:read]: List websites - get_website [mcp:read]: Get website - create_website [mcp:read mcp:write]: Create website - delete_website [mcp:read mcp:write]: Delete website - set_website_logos [mcp:read mcp:write]: Set website logos - set_website_language [mcp:read mcp:write]: Set default language - confirm_website_cname [mcp:read mcp:write]: Confirm CNAME - revoke_website_cname [mcp:read mcp:write]: Revoke CNAME - get_website_preview_access [mcp:read mcp:write]: Get preview access - rotate_website_preview_token [mcp:read mcp:write]: Rotate preview token - get_page_preview_url [mcp:read mcp:write]: Get page preview URL - list_pages [mcp:read]: List pages - get_page [mcp:read]: Get page - create_page [mcp:read mcp:write]: Create page - update_page [mcp:read mcp:write]: Update page - delete_page [mcp:read mcp:write]: Delete page - list_feeds [mcp:read]: List feeds - list_feed_items [mcp:read]: List feed items - list_deployments [mcp:read]: List deployments - deploy [mcp:read mcp:write]: Deploy website - rollback_deployment [mcp:read mcp:write]: Roll back deployment - delete_deployment [mcp:read mcp:write]: Delete deployment - list_media [mcp:read]: List media - get_media [mcp:read]: Get media - get_media_usage [mcp:read]: Get media usage - upload_media_inline [mcp:read mcp:write]: Upload small media inline - request_media_upload [mcp:read mcp:write]: Request upload URL - set_media_alternative [mcp:read mcp:write]: Set media alternative - delete_media [mcp:read mcp:write]: Delete media - get_analytics [mcp:read]: Website analytics ## Page components Page components (ordered array of { name, payload }); rendered top to bottom: - hero: { headline?: string, subheadline?: string, poster?: Media, buttonLabel?: string, buttonUrl?: URL or /path } full-width poster with overlaid headline. - text: { content: markdown } formatted body text. - split: { headline?: string, content?: markdown, textPosition?: "left"|"right", poster?: Media } image beside text. - quote: { quote: markdown, author?: string, source?: string } blockquote; the caption renders only with an author; an http(s) source URL renders as a link. - date: { date?: ISO 8601 date-time } shows a date; defaults to the page date. - media: { media: Media } image/video, linking to a full-size view. - feed-list: { feedId?: page ID, headline?: string, limit?: integer (1–2147483647), buttonLabel?: string, buttonUrl?: URL or /path } lists items of a feed page newest first; omit feedId for the current page. Omit limit for all items. Optional headline and read-more link (buttonUrl, default label Read more). - cards: { headline?: string, style?: "plain"|"boxed"|"numbered", items: [{ title?: string, content?: markdown }] } grid of cards under an optional headline: features, pricing plans ("boxed") or steps ("numbered"). - faq: { headline?: string, items: [{ question?: string, answer?: markdown }] } collapsible questions and answers under an optional headline. - cta: { headline?: string, content?: markdown, buttonLabel?: string, buttonUrl?: URL or /path } highlighted call-to-action band; the button renders only with a valid buttonUrl. Media is the componentValue of one of your media items (list_media/get_media, or the result of an upload): { id, href, mimeType, isImage, isVideo, width, height, alt }. The renderer uses these stored fields directly, so copy the whole object; the id must be your own media. A media item can have alternatives attached (set_media_alternative) that are shown instead of it while a media query matches: dark (dark colour scheme), landscape (viewport wider than tall), portrait (viewport taller than wide). Its componentValue then includes alternatives: { : Media }, which is why the whole object must be copied; a page keeps the alternatives it was saved with, so re-copy the componentValue after changing them. Text and markdown fields must be strings. Unknown component names and extra payload fields are stored and returned unchanged, but unknown components are not rendered.