Postiv AI

Create a LinkedIn carousel (multi-slide document post) from a structured CONTENT outline. For a single-canvas infographic that publishes as an image post, use create_infographic instead. Postiv generates the visual slides itself using the profile's saved brand style, colors, fonts, and slide templates — NEVER write or supply slide HTML, and don't describe pixel-level design; provide content (headlines, body, list items, stats) plus optional high-level styleInstructions. Slides are generated as editable canvas layouts: each one is a designed template with named text and image fields, so any slide's copy can be rewritten afterwards with edit_carousel_slide. Plan the cover headline at ~8 words max and slide 2 as a second hook (LinkedIn previews the cover plus a peek of slide 2). Pass planItemId when drafting a Bob plan item, so attribution and plan state stay linked. First use list_visual_templates and get_visual_template to choose and inspect a design, then pass styleId and outline.slides[].fields with exact field names and complete final copy. Template recommendedLength is guidance, not a hard limit; preserve requested details and let text wrap and grow vertically. maxLength is only an input safety cap. Generation fills the template directly without an intermediate mapping model. After creation, render_carousel prepares slide images and the PDF; show_carousel displays them only when needed. Creation itself does not display a widget. To publish, the carousel needs a companion post for the caption text (create_post), then schedule_post with carouselId.

create_carousel

Create one recurring content pillar for a LinkedIn profile (maximum five active pillars). Pillars are planning strategy, not individual post ideas. Supply a concise description and a notebook-style content definition.

create_pillar

Create an external-profile tracking list, maximum five per workspace. Check existing lists first.

create_engage_feed_list

Create a single 1080x1350 editable Fabric infographic that publishes to LinkedIn as an IMAGE post (not a document — use create_carousel for multi-slide documents). Postiv generates the visual canvas itself using the profile's saved brand style — NEVER write or supply canvas HTML; provide a structured CONTENT plan (headline of at most 8-10 words, up to 4 sections, up to 6 data points, at most 8 items total). Only use real numbers, stats, and facts the user actually supplied — never invent data. Keep total on-canvas copy to 30-60 words: pick the sections and data points that matter most rather than including everything. For a designed template, first use list_visual_templates with format infographic and get_visual_template, then pass styleId and outline.fields with exact named slots and complete final copy (sections may be empty). Template recommendedLength is guidance; preserve requested details and let text wrap and grow vertically. maxLength is only an input safety cap. Without styleId, choose one archetype or auto. Pass planItemId when drafting a Bob plan item, so attribution and plan state stay linked. Generation fills native editable text and shape objects. After creation, render_carousel prepares the PNG; show_carousel displays it only when needed. Creation itself does not display a widget. To publish, create a companion post for the caption (create_post), then schedule_post with carouselId — it publishes as an image post automatically.

create_infographic

Create a post draft in the workspace library. Follow integrationId's assignment rules. Before writing original content, search workspace knowledge and, only when a profile is selected, load its writing style with get_writing_style; skip those separate reads when the user supplied complete source/copy or a selected Bob plan item's complete drafting package already contains them. Pass templateId when using search_templates, and planItemId when drafting a Bob plan item, so attribution and plan state stay linked. The supplied copy is saved as-is; there is no automatic hook rewrite. Humanization is optional and defaults to false; pass humanize=true only when the user wants the extra humanizer pass. Drafts never publish by themselves; use schedule_post to queue publishing. Returns saved post data without UI; use show_post when presenting the finished draft for review.

create_post

Save a quick idea into the org's shared scratchpad - a mid-week buffer of notes, links, and images that decays after 30 days and feeds Bob's weekly content plan. Use for capturing a thought, article, or image worth revisiting, NOT for drafting or publishing content (use create_post for that). kind 'note' needs content; kind 'url' needs url (title/description/image are scraped server-side); kind 'image' needs imageUrl (fetched and stored server-side, max 10MB). File uploads are REST-only.

create_scratchpad_item

Create a workspace-wide writing assignment and assign it to every current member. Only owners and admins may call this. Use only when the user explicitly asks to create a team brief.

create_team_brief

Delete the real published LinkedIn comment, then clean up Postiv history. Requires an explicit deletion request and a recorded commentId belonging to this authenticated user and acting integrationId. Works for comments/replies from either source, even after untracking. Comments without a matching history record must be deleted on LinkedIn.

delete_comment

Permanently delete a scratchpad item - this cannot be undone, and also removes its stored image/file if any. Only delete when the user explicitly asks to remove a scrap; otherwise leave it alone, since every scrap already auto-expires 30 days after creation without any action needed.

delete_scratchpad_item

Deny a pending scheduled post as the API-key actor. This prevents publishing and sets the queue to approval_denied. Call only after the user explicitly denies the exact copy returned by list_pending_approvals.

deny_scheduled_post

Remove the saved image attachment from a draft post without deleting the library file or changing the post text. Does not schedule or publish.

detach_post_media

Mark a team-inbox comment as no reply needed, removing it from Postiv’s unanswered count. Explicit user action; reading alone does not dismiss. Does not reply, delete, hide or mark anything on LinkedIn.

dismiss_comment

Apply exact design edits to one canvas slide of a draft carousel or infographic, saved as ONE new version. Call get_carousel_design first and target objects by the exact id (or field name) it returned; the server never guesses, and an unknown id returns suggestions. Put every requested change for the slide in one ops batch. Ops: set (props such as fill, stroke, opacity, left, top, width, height, text, fontSize, fontFamily, fontWeight, textAlign), image (src {logo=company.com}, {asset=ASSET_ID}, an https image URL, or "" to clear), hide, show, delete, add (Textbox, Rect, Circle, Line or Image), replaceColor (every use of one colour on the slide) and order (front/back/forward/backward). External images are fetched and stored by Postiv; one that cannot be loaded fails the call before anything is saved. Returns the updated object map and new baseVersionId without UI. For visual inspection, render_carousel then show_carousel. Scheduled or published carousels are locked. Call render_carousel before scheduling a carousel.

edit_carousel_design

Edit ONE slide of a draft carousel — or the single canvas of an infographic (use slideIndex 0) — with a natural-language instruction (e.g. "shorten the headline to 5 words", "turn the list into 3 numbered steps"). Describe the change; never supply HTML or canvas JSON. New carousels and infographics are canvas layouts: the instruction rewrites the slide's named TEXT fields (get_carousel lists them), and the designed layout, colors and images are NOT changed by an instruction — use get_carousel_design and edit_carousel_design for colours, fonts, sizes, positions, images and shapes. Older HTML-based slides are re-authored as HTML instead, so visual instructions still apply to those. Saves a new version, which invalidates the previous render: call render_carousel afterwards to prepare updated exports, including the PDF required before carousel scheduling. get_carousel only reads current state and refreshes existing export URLs. Scheduled/published carousels are locked — unschedule the post first. One slide per call; takes 15-90 seconds (longer for style-heavy slides) and costs credits — don't retry blindly.

edit_carousel_slide

Edit an existing Postiv image with natural-language instructions — the way to iterate on a generate_post_image result instead of regenerating from scratch. Stacks a new version on the same image; if the image is attached to a post, the post publishes the newest version automatically. Editing takes 15-60 seconds and costs real generation credits — don't retry blindly on slow responses. Returns saved image data without UI; use show_image when presenting it for review.

edit_post_image

Generate a brand-NEW AI image for a post using Postiv's branded style presets or a freeform prompt. Only for posts that don't have the image you need yet: if a generated image already exists and the user wants it changed (different style, colors, 'more minimalist', etc.), you MUST call edit_post_image with the existing imageId instead — regenerating discards the version history and double-spends credits. The image is attached to the post (replacing any existing one), stored in the workspace image library, and published with the post by schedule_post automatically. Provide either style or freeformPrompt. Generation takes 15-60 seconds and costs real generation credits — don't retry blindly on slow responses. Returns saved image data without UI; use show_image when presenting it for review.

generate_post_image

Read cached KPI totals (impressions, engagements, engagement rate, follower growth) for one analytics account or the whole workspace, over a trailing period or an explicit start/end date range.

get_analytics_summary

Read Bob's planning settings for one LinkedIn profile: current status, posting cadence, active pillar count, and whether Bob already has a plan for the current week. Bob is Postiv's autonomous content planner.

get_bob_planning_settings

Read a carousel or infographic: current version, per-slide text, status, export readiness and fresh authorized export URLs. Does not render exports or display UI. Use render_carousel to prepare current exports, then show_carousel only when presenting them.

get_carousel

Read one canvas slide of a carousel or infographic for design editing: a compact object map (id, type, field name, text, fill/stroke colours, font, box [left, top, width, height] on the 1080x1350 canvas, hidden/locked) and the slide's colour palette, with the baseVersionId for edit_carousel_design. Use it before any design change (colours, fonts, sizes, positions, images, shapes). Match the user's words to exact object ids using the map; render_carousel then show_carousel provides optional visual inspection. Legacy HTML slides have no object map; use edit_carousel_slide for those.

get_carousel_design

Read the default Comment Plug settings for LinkedIn profiles in this workspace. Use this before schedule_post when the user expects an automatic first comment: schedule_post.commentPlug mode 'inherit' only posts a comment if this profile config is enabled, has variants, and cmmConnected is true. Pass integrationId for one profile, or omit it to inspect all profiles.

get_comment_plug_config

Read one content pillar in full: name, description, freeform content (what it covers and how to write for it), suggested posting day, mode, and 30-day performance (posts published, avg impressions, engagement rate vs. the profile average, top posts) when the pillar has published posts in that window. Use the pillar id from list_pillars.

get_pillar

Read a cached day-by-day performance time series (impressions, engagements) for one analytics account or the whole workspace. Good for spotting trends before drafting new content.

get_performance_overview

Read one post: current content, status, assigned profile, and version number. Posts with status scheduled/published are locked against edits. Returns authoritative data without opening a card; show_post presents the editor when needed.

get_post

Read cached analytics for one post. Accepts either a Postiv post id from get_post/list_posts or an analytics post id from get_top_posts. Use this for questions about a specific post instead of fetching a broad top-posts list and filtering client-side. If the post has not been published/imported yet, it returns post_analytics_not_found with the post status.

get_post_analytics

Live-refresh the complete comment thread of one of the workspace’s own posts (team source). Returns the stored post plus commentId, parentCommentId, text, author, status and unansweredCount. Reading never dismisses comments. External (feed) posts have no readable comment thread and fail with external_comments_unavailable. Incomplete refreshes fail explicitly.

get_post_comments

Read one team writing assignment in full, including its summary, writing instructions, deadline, points, creator, and every member assignment status. Use the id from list_team_briefs.

get_team_brief

Read top cached posts with full metrics and time-limited media URLs. Sort by date/engagements/impressions, or use the outlier sorts (engagement_outlier, reach_outlier) to surface posts that over-performed the account's own baseline — the best inputs for 'write more like this'. Default is all-time: pass period (trailing days) or start/end to answer time-scoped questions like 'top posts last 30 days' — do not filter client-side.

get_top_posts

Read the connected Postiv workspace (name, slug) and which scopes this API key has. Call this first to confirm you are in the right workspace and which tools are authorized.

get_workspace

Read the trained writing style for one LinkedIn profile: priority rules, voice, structure, formatting, vocabulary, custom instructions, saved learnings, and real hook/closing examples. Call this before drafting for that profile and follow modelContext as writing instructions. hasWritingStyle=false means the profile has not trained a style yet; do not invent one.

get_writing_style

Inspect a chosen carousel or infographic template before writing its copy. Returns exact field names, types, suggested copy lengths, input safety caps and illustrative examples. Request remainingSlideIndices until all layouts you will use have been inspected. Pass the same styleId to creation, with final copy in outline.slides[].fields (carousel) or outline.fields (infographic). Use show_visual_templates with styleIds and slideIndices when visual inspection is needed. No intermediate model maps copy into slots. Image fields accept HTTPS URLs. Name/avatar/CTA branding bindings come from the selected LinkedIn profile. Examples are not user facts.

get_visual_template

Deprecated name. Returns a redirect instruction without opening UI. Use show_media_upload to open the file picker, or prepare_media_upload and complete_media_upload when file bytes are accessible.

upload_media

List the LinkedIn analytics accounts tracked in this workspace, with last-import timestamps. Each account maps to one LinkedIn profile (integrationId + name). Call this FIRST before any other analytics tool — they all require an accountId, and this is where you resolve which profile the user means. Analytics data is cached (imported from LinkedIn), not live.

list_analytics_accounts

List Bob's weekly content plans for one LinkedIn profile, including every item's topic, angle, format, status, target day, and pillar plus an aggregate pillar distribution. Defaults to the current week; request up to four consecutive weeks. This is a summary only: after the user selects an item, read that exact item before drafting so its sources and writing style are loaded.

list_content_plans

List the recurring content pillars (topics a LinkedIn profile wants to be known for) that Bob's weekly plans draw from. Summary shape only — name, description, mode, and suggested posting day; call get_pillar for full content and 30-day performance. Use list_pillars -> get_pillar before drafting for a specific pillar's theme.

list_pillars

List current workspace tracking lists and external profiles. Team posts do not need tracking.

list_engage_feed_lists

Find posts to engage with. source=team (default) includes your own posts, teammates and connected company pages, from successful Postiv publications and analytics records; no tracking list required. Use status=unanswered to work through the inbox. source=feed returns posts from the workspace’s Engage tracking lists (people to engage with). Post text, likes and comment counts come from Postiv’s stored records (collectedAt says when); LinkedIn is never read for post content. Team comment threads refresh live and completely; external posts expose no comment threads. Failed posts appear in failedItems with partial=true; healthy posts remain available. If every post on a page fails, the call fails. Discovery is limited to posts already known to Postiv; this does not crawl LinkedIn for new posts. days/order/cursor use stored discovery time (scheduled time for Postiv publications). Filters apply after loading a page: empty items with nextCursor means continue. unansweredCount covers successfully refreshed team candidates only and is not a workspace total. Keep filters unchanged across pages.

list_engagement_posts

List the LinkedIn profiles and company pages connected to this workspace. Returns each profile's integration id — required by schedule_post and used by create_post/update_post to assign a post. profileType is personal (the API actor), team_member (another workspace member's personal profile), or organization_page (a company page). Call this before scheduling.

list_linkedin_profiles

List scheduled posts awaiting a decision from the API-key actor in this workspace, including exact post copy, refreshed image/video media, carousel PDF, author and target-profile details, and every approver's status and comments. It never lists another approver's private queue. Call before making any approval decision.

list_pending_approvals

List workspace posts filtered by status, assigned LinkedIn profile, or text search. ALWAYS filter deliberately instead of dumping everything: when the user asks what they have POSTED/published, use status 'published'; for upcoming content use 'scheduled' (returns only posts actually in the publish queue, each with scheduledTime); drafts are unpublished work-in-progress and must not be mixed into a 'what have I posted' answer. Each post includes its profile (id, name, headline, picture) — when the workspace has multiple profiles, group or label your summary by profile, or filter with integrationId if the user means one specific profile. Results carry a 300-char contentPreview only: before quoting, summarizing, or discussing a SPECIFIC post, call get_post for its full text.

list_posts