Lightcms

Create a new named fork workspace for staging content changes. A fork is an isolated workspace where you can edit copies of live pages without affecting the public site. When ready, an admin can merge the fork to push all changes live at once. Typical workflow: 1. create_fork — create the workspace 2. fork_page — copy pages you want to edit into the fork (returns a fork page ID) 3. update_content (with the fork page ID) — make your edits 4. merge_fork — merge all changes to the live site (admin only) Returns the fork ID needed for subsequent fork operations.

Create a new RSS/Atom import source that automatically pulls content from a feed

Create a new URL redirect. Use 301 for permanent redirects, 302 for temporary.

Create a new snippet. Snippets are Go templates that render one content item in an lc:query index page. Reference by name: <!-- lc:query filter="tag:MyTag" snippet="your-snippet-name" -->

Create a new template. Define fields with name, label, type (text, textarea, richtext, date, image, select), and options.

Create a new webhook. Returns the generated HMAC-SHA256 secret ONCE — save it immediately. Valid event types: content.create, content.update, content.publish, content.unpublish, content.delete, comment.created, content.pending_approval, asset.pending_review

Delete an approval workflow.

Delete an asset from the library. Removes both the file and database record.

Delete a collection. This does not delete the content in the collection.

Delete a discussion comment. Requires comment.delete permission (admin only).

Soft-delete a content item. The content can be restored later. Removes the static HTML page.

Delete an empty folder. Cannot delete folders that contain content or subfolders.

Permanently delete a fork and all its page copies. Deleting a merged fork removes only its history record (and any leftover copies) — the merged live pages are not affected. Requires admin role.

Delete an RSS import source

Delete a redirect.

Delete a snippet by ID.

Delete a template. Cannot delete system templates or templates that have content using them.

Permanently delete a webhook.

End the sandbox session. action='submit' keeps the fork for human review and merge (at /cm/forks/{id}); action='discard' permanently deletes the fork and all its changes. Agents cannot merge their own sandbox — merging is a human decision.

Search published content using full-text exact match, semantic (AI) similarity, or hybrid mode. Returns page titles, paths, and snippets.

Export content items with their full field data as a structured JSON array. Use for batch transformations: export → transform externally → re-import via bulk_update_content. Scope filters (all optional): - template_name: most useful for bulk workflows, e.g. "Concept Page" or "Blog Post" - category, folder_path, content_ids: narrower scoping options Use fields: ["field1", "field2"] to include only specific data fields instead of all fields. Only live pages are exported: fork copies named in content_ids come back in "skipped" as {id, reason}. Returns: total count and array of items with id,…

Admin only: force-release any lock on a content item regardless of who holds it.

Copy a live page into a fork workspace so you can edit it without affecting the public site. Provide either: - path: the URL path of the page (e.g. "/about", "/blog/my-post") — preferred when you know the URL - content_id: the MongoDB ObjectID of the content item Returns the fork page ID. Use this ID with update_content to make edits: 1. fork_page → get fork_page_id 2. update_content with id=fork_page_id to edit 3. get_content with id=fork_page_id to verify 4. merge_fork when all edits are ready (admin only) If the page is already in this fork, returns the existing fork copy.

Show whether an agent sandbox is active, and if so its fork ID, name, and the pages changed so far (with per-field diffs against live).

The audit ledger for an agent session: every content item it changed, with actions and timestamps. Defaults to the current MCP session. Requires audit permissions.

AI visibility report: requests from AI crawlers by purpose (training, AI search, user-initiated fetch) and by crawler, the pages AI reads most, visits referred by AI assistants (ChatGPT, Perplexity, Claude, Gemini…), and where they land.

Get details of a single approval request including decisions so far.

Get a single approval workflow by ID.

Get asset metadata by ID or path. Does not return file content (use the serve path to access the file).

Find all published pages that link to the given URL path. Links are tracked automatically whenever a page is published — both [[Wikilinks]] and ordinary <a href="..."> links in content fields are indexed. Use this to discover which pages reference a given page (wiki-style backlink graph), assess the impact of deleting or renaming a page, or find orphaned pages with no inbound links. Example: {"path": "/about"} returns every published page whose content contains a link to /about.

Get a collection by ID.

Get a single content item by ID or path. Returns full content including all field data (title, slug, full_path, data fields, published state). Prefer path when you know the URL: {"path": "/about"} Use id when you have the MongoDB ObjectID: {"id": "abc123"} Set include_rendered=true to also receive the fully rendered HTML output (template + theme header/footer applied). Useful for verifying what visitors see without publishing. Tip: to preview unsaved edits before publishing, use preview_content instead.

Get the current advisory lock status for a content item. Returns lock holder and expiry, or {locked: false} if unlocked.

Get a specific version of a content item with full field data.

Get the version history for a content item. Returns list of versions with timestamps.

Get a folder by ID.

Get details of a fork workspace including its status and list of pages. Returns: fork metadata + array of pages (id, title, full_path, updated_at). Use the page id with get_content or update_content to read/edit fork pages.

Per-field differences between each page in a fork and its live counterpart — the review surface for merging a fork. Status 'added' means the page exists only in the fork.

Get the status, results, and log of a specific import job

IndexNow notifies Bing, Yandex and other engines when pages change. Returns whether it is active (and why not if inactive), the key file URL, whether the key file verified, total URLs submitted, and recent submissions with HTTP results.

Get the status and results of a link check job. Status is "running", "done", or "failed".

The latest site-health scan: stale pages (no update in ~6 months), published pages missing meta descriptions, and lingering drafts. Use it as a work queue — fix issues via the sandbox workflow.

Full plain-text content of one published page by URL path, with title, description, and publish date.

The site's search & AI settings: AI crawler policy (training / AI search / user fetch, per-crawler overrides) with the resulting robots.txt, Markdown copies, default author/publisher for structured data, and feed settings.

Get site configuration including title templates.

Name, tagline, and base URL of this site, plus pointers to llms.txt and the sitemap.

Get a snippet by ID.

Get a single template by ID or slug. Returns full template including fields and HTML layout.

Get current theme settings including colors, fonts, and custom HTML for header/footer.

Get a specific version of theme settings with full data.

Get the version history for theme settings. Returns list of versions with timestamps.

Import content from CSV data. Each row becomes a content page. Specify which column is the title; all columns become content fields.

Import one or more Markdown pages into LightCMS. Each page can include YAML frontmatter (title, slug, folder, template, published, publish_at). Ideal for bulk content creation by AI agents — generate markdown with frontmatter and import directly. Example frontmatter: --- title: My Page slug: my-page folder: /blog template: Blog Post published: false --- Page body content here.

List pending approval requests. Use filter=mine to see only requests in your queue.

List all configured approval workflows. Requires approval.manage_workflows permission.

List all unique folder paths in the asset library.

List all assets in the asset library. Assets are files like images, documents, CSS, JS, etc.

List recent audit log entries. Admin only. Supports filtering by action and resource type.

List all content collections. Collections group and display content by category.

List all discussion comments on a content item.

List all content items with optional filters. Returns content metadata including title, path, publish status, and timestamps. Add include_data: true to get full field data for all items in one call, or include_fields: ["field1", "field2"] to fetch only specific fields — both avoid per-item get_content calls for bulk workflows. Up to 20 concurrent update_content calls are safe. For larger batches, prefer bulk_update_content (up to 100 items per call).

List all content folders. Folders organize content into URL path segments.

List all content fork workspaces. Forks let you stage changes to multiple pages as a batch before merging them live. Each fork shows its status (active/merged/archived), page count, and who created it. Use get_fork to see the specific pages in a fork.

List recent import jobs with their status and results

List all configured RSS/Atom import sources

All published pages on this site: title, path, and meta description.

List all URL redirects configured for the site.

List all unpublished content items that have a scheduled publish time set.

List all snippets. Snippets are reusable Go-template HTML fragments used in lc:query index page directives.

List all available templates. Templates define content structure with fields and HTML layout.

List recent delivery attempts for a webhook.

List all registered webhooks. Secrets are never returned.

Merge all pages in a fork workspace into live content. Requires admin role. For each page in the fork: - If a matching live page exists (same URL): updates it with the fork content (fork wins). If the live page was edited after the fork was created, records a conflict but still merges. - If no live page exists at that URL: creates a new live page. It keeps the publish state it had in the fork (normally draft) unless publish_new is true, which publishes it. Pages on hold always stay drafts. After merging, the fork status changes to "merged", its page copies are deleted, and the created/update…

Lock a theme version so it is protected from automatic pruning or accidental overwrite. Pinned versions are marked with locked=true in get_theme_versions. Use this to preserve milestone theme states (e.g., after a major redesign) before making further changes. Example: {"version": 5}

Render a content item's HTML without saving or publishing. Use this to verify what a page will look like before publishing. Also accepts optional title/data overrides to preview unsaved edits: {"id": "abc123", "data": {"body": "<p>New text</p>"}} Returns rendered_html and any warnings (missing required fields, unclosed tags, unresolved placeholders).

Publish a content item, making it visible on the public site. Generates the static HTML page.

Publish multiple content items in a single call. Use this instead of calling publish_content in a loop. Examples: - Publish specific pages: {"ids": ["abc123", "def456"]} - Publish all drafts at once: {"publish_all_drafts": true} Held pages and fork copies are never published by this tool: they come back in "skipped" as {id, reason}. publish_all_drafts ignores fork copies entirely. Returns published IDs, skipped items, and any failures.

Delete the page copies still attached to a merged or archived fork. Requires admin role. Merges clean up after themselves since v7.4; forks merged before that kept their copies. Live pages are never touched, and the fork record is kept. Active forks are refused. Use dry_run: true first to see the count and the list of {id, full_path}. Returns the number deleted and writes an audit log entry.

Regenerate all published static HTML pages. Use after major theme or template changes.

Generate a new HMAC-SHA256 signing secret for a webhook. The old secret stops working immediately. The new secret is returned ONCE — save it.

Regenerate vector embeddings for all published content. Required after initial setup or if embeddings become stale.

Reject an approval request. A comment explaining the rejection is required and will be posted to the content's discussion thread.

Release the advisory lock on a content item held by the current API user.

Remove a page from a fork workspace (discards the fork copy, does not affect the live page).

Repair damage the admin editor could do when saving a fork copy before v7.4.1: (1) fork copies left flagged published get the flag cleared; (2) live published pages whose generated HTML file is missing are regenerated. Page content is not changed: no versions, no IndexNow pings. A flagged copy with has_live_page=false is a page created inside its fork, where the flag meant publish-on-merge — after the repair, merge that fork with publish_new to publish it. Run with dry_run first and show the user the lists. Requires admin role.

Restore a soft-deleted content item. Regenerates static page if content was published.

Revert theme to a previous version. Creates a new version with the old data.

Revert content to a previous version. Creates a new version with the old data.

Undo every content change made by an agent session as a unit: content the session created is soft-deleted, deleted content is restored, and updated content reverts to its latest pre-session version. Defaults to the current MCP session ("undo everything I did"). Non-content changes (templates, settings) are listed for manual review but not auto-reverted.

Run a site-health scan now and return the fresh report. Scans also run automatically once a day.

Set a future publish date/time for a content item. The scheduler automatically publishes it when the time arrives.

Execute a search-and-replace limited to a subset of pages. ALWAYS run scoped_search_replace_preview first and show results to the user before executing. Scope options (all optional): - content_ids, folder_path, template_name, category Set auto_republish: true to immediately re-publish all previously-published pages after updating them, collapsing the execute + publish_multiple flow into one call. Only live pages are changed: fork copies named in content_ids come back in "skipped" as {id, reason}. Example: {"search": "old text", "replace": "new text", "folder_path": "/blog", "auto_republish…

Preview a search-and-replace limited to a subset of pages. Safer than site-wide replacement. Scope options (all optional — leave blank to match all pages): - content_ids: specific page IDs - folder_path: pages under /blog, /docs, etc. - template_name: pages using "Concept Page", "Blog Post", etc. - category: pages with a matching category Only live pages are searched: fork copies named in content_ids come back in "skipped" as {id, reason}. Example: {"search": "old text", "replace": "new text", "folder_path": "/blog"} Always run preview before execute.

Search across all content items by title or full text. Returns matching content with paths and match context.

Execute a site-wide search-and-replace across all content. Modifies every matching page permanently. MANDATORY workflow: 1. Run search_replace_preview and show the user which pages will be affected. 2. Get explicit user confirmation before executing. 3. Run search_replace_execute with a clear version_comment. Set auto_republish: true to immediately re-publish all previously-published pages after updating them, collapsing the execute + publish_multiple flow into one call. For targeted replacements, use scoped_search_replace_execute instead.

Preview a site-wide search-and-replace without making any changes. ALWAYS run this before search_replace_execute. Returns: affected page count, total match count, and per-page field breakdown. Only live pages are searched; fork copies are never matched or rewritten. For targeted replacements (a folder, template, or category), use scoped_search_replace_preview instead.

Full-text search across the published site. Returns matching pages with snippets.

Turn automatic IndexNow submission on or off for this site (on by default).

Start a sandboxed editing session backed by a content fork ("pull request for content"). While the sandbox is active: - update_content / update_content_by_path transparently copy pages into the fork and edit the copies (copy-on-write). Live content is never modified. - create_content creates new pages inside the fork. - Publishing, deleting, bulk operations, and search/replace-execute are blocked. - The site is unaffected until a human reviews the fork at /cm/forks and merges it. Recommended for any multi-page or risky editing task. End with end_agent_sandbox.

Start an asynchronous broken-link check across all published content. Returns a job_id immediately. Poll get_link_check_results(job_id) until status is "done". Example workflow: start_link_check() → {job_id: "abc123"} get_link_check_results("abc123") → if status=="running", wait and retry; if "done", review broken_links

Explicitly submit a content item for editorial approval. Contributors must submit content for approval before it can be published.

Submit pages to IndexNow now. Changes are already submitted automatically on publish/update/delete; use this to resubmit specific paths, or all=true to submit every published page (once per hour max).

Manually trigger an RSS import source to run immediately

Remove the lock from a previously pinned theme version. Example: {"version": 5}

Unpublish a content item, removing it from the public site. Removes the static HTML page.

Update an existing approval workflow.

Update a collection's settings.

Update an existing content item by ID. Creates a new version automatically. Only send fields you want to change. For partial data updates, only the keys you include in "data" are changed — existing keys are preserved (merge semantics). Use clear_fields: ["field1", "field2"] to explicitly set fields to empty string. Set dry_run: true to validate the update without saving. To update by URL path instead of ID, use update_content_by_path. Always include version_comment so the version history is useful. Up to 20 concurrent update_content calls are safe. For larger batches (>20 items), prefer bulk…

Update content identified by its URL path instead of its ID. Useful when you know the page URL but not the MongoDB ID. Example: {"path": "/about", "title": "About Us", "data": {"body": "<p>Updated content</p>"}} The path always resolves to the live page, never to a fork copy (inside an agent sandbox the write goes to the sandbox copy as usual). Only the fields you provide are changed. Always include a version_comment describing what changed.

Update an RSS import source configuration

Update an existing redirect.

Change search & AI settings. Only the fields you pass change. Blocking AI training crawlers does not affect Google/Bing search.

Update site configuration. Title templates support {{title}} and {{site_name}} placeholders.

Update an existing snippet's name and/or HTML template.

Update an existing template. Changing the HTML layout will regenerate all content using this template.

Update theme settings. Only the fields you provide are changed — all other settings are preserved (partial update, safe to call without get_theme first). Changing header_html or footer_html triggers background regeneration of all published pages. Changing colors, fonts, or custom_css does NOT require content regeneration. Use pin_theme_version to protect important milestones before making major changes.

Update an existing webhook (name, URL, events, active). Partial update — only provided fields change.

Upload or replace an asset in the asset library. Re-uploading to the same serve_path replaces the existing file in place — no need to delete first. Provide file content via one of: - file_path: Absolute local path to the file (preferred for files >100KB — avoids MCP transport size limits) - data_base64: Base64-encoded file content (fine for small files) Validates file type and MIME type for security.

Fetch a public URL and store it as a LightCMS asset. Useful for importing images or files from the web without downloading them locally first. Example: {"url": "https://example.com/logo.png", "serve_path": "/assets/logo.png", "description": "Site logo"} If serve_path is omitted, the filename is derived from the URL. Returns id, serve_path, mime_type, and size.