WebsitePublisher

Manage MAPI entities (schema level — like database tables). Single entry point — pick an operation. list = all entities in the project. create = entity_name(+properties[], description?, plural?, public_read?). update = entity_name + any of description/plural/public_read/policy_json. delete = entity_name — irreversible, drops all data. schema = entity_name → full property list + current policy. add_property = entity_name+property_name+type(+length?,required?,default?,in_listing?) — add a column to an existing entity. delete_property = entity_name+property_name+confirm:true — irreversible, drops the column and its data.

entities

Manage reusable HTML fragments (shared nav, footer, banners) embedded via <!--#wps-include fragment="name" -->. Single entry point — pick an operation. list = all fragments + versions. create = name+content. update = name+content, full replace (supports base_version_hash optimistic lock + force). patch = name+patches, targeted find/replace, efficient for small edits, preserves version history. delete = name. versions = name → version history. rollback = name + target_version|target_version_hash. All changes invalidate cache automatically.

fragments

Returns agent skill documents for building WebsitePublisher sites: page-structure patterns, code snippets and guidelines. "main" covers site-building and the go-live checklist, "design" covers typography, color, layout and animation, "dev" covers internal platform conventions. The main skill is served in sections: a plain call returns the rules that always apply plus an index of section slugs, and you fetch the rest with the section parameter as you need it.

get_skill

Read the full content of an existing asset. Text assets (js, css, json, svg, txt, html) are returned as UTF-8 strings in `content` with `encoding: "utf-8"`. Binary assets (images, fonts, pdf) are returned as base64 in `content` with `encoding: "base64"`, but only if size ≤ 1MB; above that the content is omitted and the `url` field can be used to download directly. Always returns `version_hash` (first 8 hex chars of the SHA-256 of the content) for optimistic concurrency on subsequent updates.

get_asset

Reads back what the user changed during a visual edit session. Returns a structured changelog with uploaded images, dimensions, style changes, etc. Use this after sharing the edit URL with the user to see what they did.

get_edit_session_changes

Get the full schema of a specific integration: all available endpoints, required fields, and input parameters — plus its usage guidance (how to apply, pages to build, anti-patterns) when available. Full catalog of every platform building block, no auth needed: https://www.websitepublisher.ai/integrations.txt.

get_integration_schema

Get a specific page with its content. Returns version and version_hash for optimistic-concurrency checks on subsequent updates.

get_page

Get the version history of a page. Returns metadata (version numbers, hashes, timestamps, change summaries) — no content. Useful to check what changed or find a version to rollback to.

get_page_versions

Get detailed status of a specific project including page, asset, and entity counts

get_project_status

Use first when the person wants to build, change or publish a site, page, shop or app. No arguments. Returns the person's projects with their live URLs, which project to build in, and the exact next call that puts a page live. Read-only.

get_started

Get the current tracking scripts configured for a project. Returns head_scripts and body_scripts, or null if none configured.

get_tracking_scripts

Get visitor analytics for a project. Returns pageview counts, unique visitors, top pages, referrers, device breakdown, UTM data, or daily trend. Use period "today", "7d", "30d", or "90d".

get_analytics

Get the current visitor auth configuration for a project. Shows whether auth is enabled, which methods are allowed, and the success redirect URL.

get_visitor_auth_config

List all assets (images, CSS, JS, etc.) in a project

list_assets

List all configured forms for a project. Shows form names, required fields, configured actions, and submit limits.

list_forms

List every integration for a project and whether it is configured (vault secrets present, ready to call) or still needs setup. Compact by default: id, name, category and endpoint names. For the input fields of one integration call get_integration_schema — that also returns its usage guidance. Use detail:true only when you really need every endpoint description of every integration (large response). Full catalog of every platform building block, no auth needed: https://www.websitepublisher.ai/integrations.txt.

list_integrations

List the pages in a project as a compact index: slug, title, language, landingpage, version, version_hash, created_at, updated_at. No content by default — call get_page for the one page you need. Paginated (default 50, max 100 per call); the response has pagination.total and a hint when more pages exist. Load the index once per session and reuse it instead of calling this again before every edit. Fragments are not included (use the fragments tool).

list_pages

List all projects the authenticated user has access to. Returns the id, name and subdomain of each project. At the start of a conversation prefer get_started: it includes this list plus the project to build in and the next call.

list_projects

List all scheduled tasks for a project, showing their status, next run time, and last execution. A task with disabled_by_system=true was switched off by the platform after repeated failures: disabled_reason says why. Fix the cause named there and switch the task back on — do not create a second task alongside it.

list_scheduled_tasks

List all secrets stored in a project's vault. Returns metadata only (key names, prefixes, service types, status) — never the actual secret values. Use this to check what credentials are stored before setting up integrations.

vault_list_secrets

Apply targeted text changes to an existing asset (js, css, json, svg) without re-uploading the full file. Uses find/replace on the file content. Binary assets (images, fonts) are updated by re-uploading with overwrite: true instead.

patch_asset

Apply targeted changes to an existing page without sending the full content. base_version_hash is optional — if omitted, find/replace acts as the natural conflict guard. Supports replace, delete, insert_before and insert_after operations. Much more efficient than a full replacement for small edits.

patch_page

Manage MAPI records (data level — rows within an entity). Single entry point — pick an operation. list = entity_name(+page?,per_page?,sort_by?,sort_order?). get = entity_name+record_id. create = entity_name+data. update = entity_name+record_id+data, partial — only provided fields change. delete = entity_name+record_id — irreversible.

records

Remove a form configuration from a project. Visitors will no longer be able to submit this form.

remove_form

Remove an integration by permanently deleting all its vault secrets. After removal, the integration endpoints will no longer work until it is reconfigured. Full catalog of every platform building block, no auth needed: https://www.websitepublisher.ai/integrations.txt.

remove_integration

Remove all tracking scripts from a project. Pages will no longer have tracking code injected.

remove_tracking_scripts

Rollback a page to a previous version. Creates a new version with the old content (audit trail preserved). Specify either target_version (number) or target_version_hash. Available versions are listed in the page version history.

rollback_page

Execute an integration action — e.g., send an email via Resend, create a payment via Mollie. Vault credentials are resolved server-side, so API keys are never handled directly. Non-built-in integrations must be configured before use. The exact endpoint name and required input fields are available from the integration schema. Payment endpoints only create checkout sessions hosted by the payment provider (e.g. Mollie); card data is entered on the provider-hosted checkout page and never passes through or is stored by WebsitePublisher.

execute_integration

Search every capability the platform has, across ALL projects this user can reach, in plain language — "email archive", "search my inbox", "send invoices", "book a table". Returns real, live capabilities from the integration manifests, including the project_id to call them on, so you never need to know the project up front. ALWAYS call this before telling the user something is not possible, and before reaching for an external tool, web search or a third-party connector: the platform ships 100+ integrations and the one you need is usually already there. Returns the exact service and endpoint names to pass to execute_integration. Full public catalog as one plain-text file, no auth or MCP needed: https://www.websitepublisher.ai/integrations.txt (one line per integration: slug, category, endpoint names).

search_integrations

Set Google Analytics, Google Tag Manager, Meta Pixel, or other tracking/conversion scripts for a project. Scripts are automatically injected into every page: head_scripts before </head> (for analytics/GTM), body_scripts before </body> (for conversion pixels). Set a field to null or omit it to clear.

set_tracking_scripts

Shared task tracker (TAPI) with append-only history. Single entry point — pick an operation. list = the working overview — compact by default: active tasks (open/in_progress/blocked), roots only, most recently updated first, max 50 rows, no descriptions; verbose:true for the full payload. get = slug|id → single task + architecture versions. history = slug|id → full history, with_content=true fetches MD from S3. create = slug+title(+description, parent_id, related_to, tags, scope). add_history = slug|id+type+summary — the primary way to report progress, notes, snippets, blockers and architecture docs; supports snoozed_until. update = slug|id + any of title/description/parent_id/related_to/tags — metadata only, logs an audit note; status and completion stay history-driven via add_history. delete = slug|id+confirm:true — soft delete, refused while the task has children. search = AND-match on distinctive words across title/description/slug + history summaries. export = full Markdown archive of all tasks, returned as a short-lived download link — not for getting an overview; use list for that.

tasks

Replace an existing page with full new content. Changes are live immediately. Requires base_version_hash (the current page version) to prevent accidental overwrites — send force: true to skip this check. Full replacement rewrites the entire page; small targeted edits are more efficient via a partial update and preserve version history.

update_page

Upload an asset (image, font, PDF, etc). Provide exactly one of: content (base64), content_text (plain text for JS/CSS/JSON/SVG — preferred, saves tokens), or source_url (public HTTPS URL for images). Set overwrite: true to replace an existing asset.

upload_asset