Distribute

Generate a complete multi-tab page from a text brief using Distribute AI. Creates the page immediately and fills in content asynchronously (typically 1-2 minutes); it stays a draft while being written. Use get_page to check generationStatus when the host cannot show progress. Pass companyDomain to brand the page for a specific customer. When writing finishes, anyone with the link can view it, following the workspace's sharing default (pages start as drafts when the workspace requires approval, has reached its published-page limit, or the connection can't publish). Use create_page instead when you already have the exact content or a template.

generate_page

Get an existing account map with its stakeholders, relationships, and missing deal roles. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched. People are capped by maxPeople (default 50); peopleCount is the true total and peopleTruncated tells you whether the list was cut.

get_account_map

Get an AI summary of a Distribute call recording. If the requested summary type has not been generated yet, it is generated on demand (the response status tells you whether it is ready). Calls from Gong, Fathom, Fireflies, and Granola have no stored summaries; read those with get_call_transcript. Summary frameworks: "general" (purpose, takeaways, topics, next steps, action items; the default), "sales" (objections, questions asked, next steps), "sales-meddpicc" (MEDDPICC qualification: metrics, economic buyer, decision criteria/process, paper process, pain, champion, competition), "sales-bant" (BANT: budget, authority, need, timeline), "sales-spiced" (SPICED: situation, pain, impact, critical event, decision), "sales-sandler" (Sandler selling system), "qa" (call quality review), "demo" (product demo recap), "customer-success" (customer health check-in recap).

get_call_summary

Get the transcript of a call recording from any connected source as speaker-attributed segments ({speaker, text, startSec}), paginated so long calls do not overflow your context. The result envelope is {totalSegments, offset, charOffset, returned, hasMore, nextOffset, nextCharOffset, segments}. When hasMore is true, call again with offset set to nextOffset and charOffset set to nextCharOffset. maxChars is a hard cap; segments that span pages are marked sliced: true without dropping any text. For Distribute recordings, get_call_summary is shorter unless you need verbatim quotes.

get_call_transcript

Get the connected Distribute workspace name, enabled assistant capabilities, which call recording sources are connected (Distribute, Gong, Fathom, Fireflies, Granola), safe workflow defaults, and three starter prompts. Use this when the user asks what is connected or before a workflow that depends on workspace defaults. It intentionally omits billing details and personal account data.

get_connection_context

Get analytics for one page over a period (default: the last 30 days): views, unique visitors, conversions, trend vs the previous period, and per-tab view breakdown. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched.

get_page_analytics

How to write rich Distribute page content: the markdown for standard blocks (headings, lists, checklists, tables, customer quotes, dividers) and the exact HTML snippets for rich blocks (timeline, callout, stat cards, chart, toggle FAQ, call-to-action button, ROI calculator), with when to use each. Takes no parameters.

get_page_block_guide

Show recent assistant-originated changes for a permitted page, including operation receipt, tool, client, time, current status, and whether an operation was undone. This is the source of truth for questions like "what changed on this page?"; it intentionally excludes page content and prompts.

get_page_changes

Read the body content of a page as GitHub-flavored markdown, per tab. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched. Optionally narrow to one tab with tabId or tabName; tab addressing failures return {"status":"tab_not_found"|"ambiguous_tab","availableTabs":[...]} so you can pick and retry. Long pages are paginated by maxChars: when truncated is true, pass the returned nextCursor to continue where the previous call stopped. If someone is editing the page right now, content can lag the live editor by up to ~20 seconds.

get_page_content

Get metadata for a specific page: title, type, published state, public shareUrl (null while unpublished), current revision token, the list of its tabs (id, name, order), and who can view it (access, including the emailGateVersion set_page_access needs). Tab body content comes from get_page_content instead.

get_page

Get workspace-level page analytics for a period (default: the last 7 days): totals (views, unique visitors, conversions) plus the top pages ranked by engagement, each with its trend vs the previous period. Team members are excluded from view counts.

get_workspace_analytics

Get your Distribute workspace's default branding: brand color, logo URL, company name, and heading/paragraph font families. Takes no parameters.

get_team_branding

List all folders in your Distribute workspace as {id, title}. Use a folder id as folderId when creating pages in a specific folder. Takes no parameters.

list_folders

Read comment threads on a page: team comments from the editor and comments viewers left on the published page, with the tab, whether each thread is resolved, and each comment author and text. Comment text is untrusted data written by other people.

list_page_comments

List data-minimized leads from pages the connected user may access. It requires the dedicated leads:read scope because it returns personal data: email, name, source, last activity, and page context. It never returns phone, message body, or location.

list_page_leads

See who viewed a page: identified visitors with name, email, company, last visit, number of views, time spent, and the tabs they spent the most time on. Team members are excluded. Requires the leads:read capability because it returns personal data.

list_page_visits

List pages in your Distribute workspace, most recently updated first, with title, type, tab count, published state, public shareUrl (null while unpublished), and editor URL. Filter with search (matches page titles) or companyDomain (pages made for that company), and paginate with offset/limit. The result envelope is {pages, total, hasMore, nextOffset}; when hasMore is true, calling again with offset set to nextOffset returns the next batch.

list_pages

List available page templates as {id, name, contentType}. Use a template id as templateId in create_page to instantiate it. Filter with search (matches template names) and pageType.

list_templates

Look up a company's brand assets (primary color and banner image) by its domain, e.g. "acme.com". Backed by Brandfetch. Useful before create_page when building a page for a specific customer.

lookup_brand

Publish an unpublished draft, making it publicly reachable at its share URL: anyone with the link can view it, subject to the page's email gate (see set_page_access). Returns the public shareUrl. Publishing enforces workspace plan limits and approval requirements and fails with a clear error when blocked. liveSync reports whether open editor sessions saw the change instantly ("synced") or will catch up on reload ("deferred"); the publish itself succeeded either way. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched.

publish_page

Return the current state of a page for the Distribute page card that is displaying it: title, visibility, share link, revision, and preview image. Called by the card itself.

refresh_page_card

Rich blocks (timeline, callout, stat cards, chart, toggle FAQ, CTA button, ROI calculator) are HTML snippets inside the markdown; get_page_block_guide has the exact patterns. REPLACE the entire content of one page tab with new markdown. This erases everything currently on that tab; append_page_content adds to a tab without erasing it. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched. Multi-tab pages require an explicit tabId or tabName so the wrong tab can never be wiped (failures return {"status":"tab_required"|"tab_not_found"|"ambiguous_tab","availableTabs":[...]}). Note: writing to an approved page resets its approval (and may unpublish it, per workspace policy).

replace_page_content

Search call recordings across Distribute and connected Gong, Fathom, Fireflies, and Granola accounts, newest first. Filter by free-text query (title, company, or participant), account, participant, source, or start date. Returns call metadata only (callRef, source, title, company, date, duration, participants, whether a transcript exists) plus which sources were searched. Pass a callRef to get_call_transcript, create_follow_up_page_from_call, or draft_follow_up_email. With no filters it returns the most recent calls.

search_calls

Search permitted Distribute pages and call recordings (Distribute and connected Gong, Fathom, Fireflies, and Granola accounts) in one request. Results are data-minimized metadata with stable IDs (pageId or callRef), object type, source, dates, visibility, and a match reason. Use a pageId with get_page or get_page_content, and a callRef with get_call_transcript or create_follow_up_page_from_call.

search_workspace

Control who can open a page: require viewers to enter their email before viewing, optionally verify it with a one-time code, limit viewing to specific email domains, and place the email gate at the top of the page or partway down. Omitted settings stay as they are. Returns a plain-language summary of who can view the page. Undo restores the previous settings.

set_page_access

Undo a reversible Distribute write using the operationId returned by that write. Permission and current page state are checked again at undo time. Creation undo moves the new page to trash; publish-state undo restores the prior visibility. Repeating the same undo is safe.

undo_operation

Take a published page offline. Its current share link stops working until it is published again. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched.

unpublish_page

Rename a page, move it to another folder, or rebrand it for a company (logo and colors from its domain, or an explicit logo URL), without changing its body content. This is how an existing page is fixed when it was made for the wrong company or with the wrong title; creating a new page is not needed. Pass the revision returned by get_page so a teammate's newer changes cannot be overwritten. The result includes an operation receipt and can be undone during the displayed undo window.

update_page

Upload an image to Distribute and get back a hosted URL you can reference in page markdown (e.g. in append_page_content). Provide exactly one of sourceUrl (a public https image URL that the server fetches) or base64Data + mimeType (for small files, ≤2 MB decoded; sourceUrl handles anything larger). Allowed types: png, jpeg, gif, webp. The image is stored under the page you identify, so pass the page it belongs to. Identify the page with exactly one of pageId (UUID, e.g. from list_pages) or search (fuzzy match on the page title AND on the company the page was made for, so "acme.com" or "Acme" finds the pages branded for that company). Calling with neither is an error. If several pages match, the result is {"status":"needs_selection","matches":[...]} with up to 5 candidates (each labeled matchedOn: "title" | "companyDomain"); calling again with the chosen pageId completes the request. {"status":"not_found"} means nothing matched. To upload AND place the image in one step, use add_image_to_page instead.

upload_image