Get the account. Identity, plan, the calling key and its scope, and the configured webhook endpoints. The cheapest call to prove a key works.

Get the account overview. The numbers behind the portal's Overview page, computed by the same data layer: a rolling window against the same length of time before it (current/previous pairs). Messages per channel, AI vs automation replies, flow entries (with an attached comment trigger's fires on the flow's own row so a comment is never counted twice), standalone comment trigger fires, contacts, appointments with the next few, knowledge gaps by status, what needs a human, API sends, and plan usage for the billing cycle. Usage is always the billing cycle, whatever the range.

List contacts. Cursor-paginated, newest first. Filters combine with AND.

Get a contact. One contact by id.

Create or update a contact. Upsert keyed by (channel, platform_user_id). 201 when created, 200 when an existing contact was updated. Fields you omit on an existing contact are left untouched, tags included.

Update a contact. Partial update. Only the fields you send change. An empty string clears name, email, phone or notes. custom_fields merges per key; null clears one value.

List tags. Every tag definition, ordered by name. Not paginated.

Create a tag. Idempotent on name: 201 when created, 200 when it already existed (colour unchanged).

Attach a tag. Attach by name. The tag is created if it does not exist. Re-attaching is a no-op.

Detach a tag. Detach by tag id or URL-encoded name. Detaching a tag the contact does not have is a 200.

Delete a tag. Removes the definition and detaches it from every contact and conversation.

List custom field definitions. Active fields by default.

Create a custom field. field_key is permanent after creation. Up to 15 active fields.

Rename or archive a custom field. Change the label, or archive and unarchive. Archived fields keep their values but stop appearing on contacts.

List conversations. Cursor-paginated, newest first. Deleted conversations are never returned.

Get a conversation. One conversation by id.

List messages in a conversation. The transcript, newest first. Purged bodies appear with text null and body_purged true.

Update a conversation. Two optional fields. status: active, resolved or archived; resolving or archiving hands the thread back to the AI, like the inbox. ai_paused: true takes the thread away from the AI as the account owner (the inbox Claim button, reopens a resolved thread); false hands it back (the inbox Transfer to AI). Resume is refused with 409 when the pause came from a solicitation pause, a fraud flag or a connected-account pause; those are lifted in the portal. status is applied first, then ai_paused.

Send a message. Send text, a WhatsApp template, a reaction, or a typing indicator on any channel. Route by conversation_id, by platform plus platform_user_id, or by to_phone for SMS and WhatsApp. SMS, WhatsApp and Telegram can start a conversation cold; Facebook and Instagram need an inbound message first. idempotency_key is required in the body, or send the Idempotency-Key header. This reaches a real person on their phone. Confirm the recipient and the text with the user before calling.

Assign or unassign a conversation. Body {"assigned_to": } to assign, {"assigned_to": null} to unassign. Moves pending or escalated to active and adds a system line to the transcript, exactly like the inbox.

List team members. Everyone a conversation can be assigned to. Not paginated. channels is null when the member can see every channel. Use it to find the member id for assign_conversation.

List blocked users. Everyone the AI is currently refusing to answer. Not paginated.

Block a user. Send exactly one of ig_username (blocks by Instagram handle, even before they have messaged) or conversation_id (blocks the person behind an existing conversation on any Meta channel).

Unblock a user. The AI answers their next message. The block row is kept for audit, so the id is not reusable.

List opt-outs. Who has opted out, on which channel, when, and how. starting_after is an opaque cursor here, not an id.

List appointments. In-chat bookings, newest first.

Get availability. The open slots for one day, from business hours, the service duration, the buffer, the minimum notice and existing bookings. What the bot and the portal picker offer. Empty on a closed day.

Get an appointment. One appointment by id.

Update an appointment. Edit details and/or status, the portal's edit form. Send only what changes. Detail edits are pushed to the connected calendars; a status change to cancelled removes the calendar event and fires appointment.cancelled. To move the time use reschedule, which checks the slot. Edits are pushed to the connected calendar, and a status of cancelled removes the event. To move the time, use reschedule_appointment.

Book an appointment. Books a slot the way the portal calendar does: the account's required customer fields are enforced (422 missing_fields), the time must be in the future and free (409 slot_unavailable), the appointment is confirmed at once, synced to Google Calendar and GoHighLevel when connected, and appointment.booked fires. Attach conversation_id to book for a person you are chatting with; they get the confirmation DM unless send_confirmation is false. date and time are the customer's wall-clock time in timezone (the account timezone when omitted).

Reschedule an appointment. Moves a pending or confirmed appointment to a new date and time after checking the slot is free, updates the connected calendars, fires appointment.rescheduled, and DMs the customer when the booking came from a conversation. 409 slot_unavailable when the slot is taken, 409 not_active when the appointment is not pending or confirmed.

Cancel an appointment. Sets status cancelled, removes the Google Calendar and GoHighLevel events, fires appointment.cancelled. 409 already_cancelled if it already is. Send {} as the body.

List flows. Automation flows with lifetime stats. The canvas definition is not exposed.

Get a flow. One flow by id.

Get a flow definition. The definition alone: the draft when one exists, else the published version.

Get the node catalogue. Everything needed to build a flow: every trigger source (channels it can fire on, whether it needs keywords, its config), every node type (input and output ports, config fields and allowed values, which are terminal), the enums, the placeholders, the structural rules, and this account's live options: custom fields, tags, bookable services, flows (for goto_flow), comment triggers (for linked comment flows), the plan's node and flow caps. Read it before composing a definition. Read this before writing a flow definition. It lists every node type, its config fields and output ports.

Validate a flow definition. Dry run of the same validation create and update apply. Nothing is saved. Validate a definition without saving it. Loop on errors[] until it is clean, then call create_flow.

Create a flow. Creates a DRAFT from a definition (see GET /flows/nodes for the catalogue). The definition is validated first: structure, per-node config, the plan's node cap, channel compatibility with the trigger source, keyword collisions with active flows, and references to custom fields, comment triggers and other flows (422 invalid_definition with errors[] and warnings[]). Nothing runs until you publish and activate. The visual builder lays the nodes out the first time the flow is opened.

Update a flow. Rename, retarget channels, set the tag, and/or replace the definition. A new definition is saved as the DRAFT and validated the same way as on create; the live version keeps running until you publish. Positions are dropped when the definition changes (the builder re-lays it out).

Publish a flow draft. Makes the draft the live definition (new version). Runs the full validation and keyword-conflict check (422 invalid_definition). Publishing a flow that was never live also activates it when the plan has a free live-flow slot (403 plan_limit_reached otherwise). Fires flow.published, and flow.activated when it went live. Send {} as the body.

Activate a flow. Sets the flow live. Publishes the draft first when there is one, exactly like the builder's toggle. Plans cap live flows (403 plan_limit_reached). Send {} as the body.

Pause a flow. Stops the flow and cancels its open sessions. A linked comment trigger is paused with it. Fires flow.paused. Send {} as the body.

Duplicate a flow. Copies the flow as a new draft. Send {} as the body.

Discard a flow draft. Throws away the unpublished draft; the live version stands. Send {} as the body.

Delete a flow. Deletes the flow, its versions, sessions and media. A linked comment trigger is detached and paused. Permanent: the flow, its versions, sessions and analytics are gone. Confirm with the user, and prefer pause_flow.

Get flow analytics. The builder's Analytics panel for one flow: runs, completions, completion rate and median time for the window, the same numbers for the preceding window (comparison), lifetime totals, outcomes (emails and phones collected, leads marked, bookings, link clicks, handoffs), per-node reach, continue, response and loss rates, and per-port branch splits. Analytics are per published version: version defaults to the current one; versions lists the others. Only sessions with tracking are counted; other_version_runs_excluded says how many ran under another version in the window.

List flow sessions. One session per person per run of a flow. Attribution carries post_id, comment_id and matched_keyword only, never what anyone typed.

Get a flow session. One session by id.

List events for a session. The ordered path one person took through a flow, oldest first.

List flow events across sessions. Every session event in one feed, oldest first. Ids only increase, so starting_after is a durable high-water mark for a sync loop. Use flow_session, not session_id, to filter by session.

List comment triggers. Standalone comment triggers, newest first, each with the flows linked to it. A flow whose own trigger node carries keywords (native) is not a trigger and appears only on /flows as a comment source with mode native.

Get a comment trigger. One trigger by id with its linked flows.

Simulate a comment. Dry run: what would happen if this text were commented on this channel and post. Runs the live matcher over every trigger in id order and reports the first that fires, the reason each other one did not, the linked flow that would start, or the native flow that would answer if no trigger fires. Nothing is sent and nothing is recorded. Per-person rules (opt-out, blocklist, once-per-user) are not simulated. Dry run, nothing is sent. Use it to check a trigger matches before and after changing it.

Create a comment trigger. Creates a trigger. It goes live at once unless the plan's live-trigger cap is reached, in which case it is saved paused and created_paused is true (never refused). Buttons are optional, up to three. A reply button is still worth adding: its tap opens the conversation so follow-ups can be sent. Story and Live triggers are Instagram only and never post-scoped. Send draft: true to save an incomplete trigger paused, skipping the template and button requirements. It goes live at once and starts DMing real commenters. Confirm with the user first, or send draft: true to save it paused. A paused trigger stays paused through edits until you send active: true.

Update a comment trigger. Send only what should change; the rest is kept and the whole trigger is re-validated. active: true activates (refused with plan_limit_reached at the live cap, or flow_conflict if the linked flows are not ready); active: false pauses and pauses the linked flows. Changing button titles or types on a trigger with linked flows pauses those flows and the trigger for branch review; the response carries flow_impact. Edits a trigger that may be live. The whole trigger is re-validated, so a draft needs its DM text and a button before any edit saves. It never switches a trigger on unless you send active: true; confirm with the user before you do.

Duplicate a comment trigger. Copies every setting into a new trigger named "(Copy)", paused, with fresh stats. Activate it with PATCH active: true. Send Content-Type: application/json with an empty object body; the server refuses a bodiless POST.

Delete a comment trigger. Soft delete: the trigger stops firing and disappears from lists, its stats are kept, and any linked flows are detached and paused (detached_flows says how many).

Get comment trigger stats. Lifetime counters for one trigger: fires, public replies, DMs (and how many were AI-written), link clicks, unique people, split by channel.

List recent posts. Recent posts and Reels on the connected Instagram or Facebook account, newest first, straight from Meta. Use an id as post_id when creating a trigger scoped to one post. 409 not_connected when that channel is not connected.

List handled comments. Comments the AI Comment Manager handled, newest first, kept 30 days. Each carries the comment, its sentiment, what moderation did, the AI draft, and what happened to the reply. status=pending is the approval queue.

Get a handled comment. One log row by id.

Get comment manager stats. How many replies are waiting for approval, and today's and this month's handled, replied and moderated counts against the plan's caps.

Approve a reply. Posts the pending reply publicly under the comment, as the account. Send {"text": "..."} to post an edited reply instead of the AI draft; the draft is kept and edited becomes true. Only a pending comment can be approved (409 not_pending). A successful send counts against the plan's reply cap. Send Content-Type: application/json with an empty object when not editing; the server refuses a bodiless POST.

Reject a reply. Drops the pending draft. Nothing is posted. Only a pending comment can be rejected (409 not_pending). Send an empty JSON object as the body.

Bulk approve replies. Sends the AI draft for each pending id, in order. Ids that are not pending or fail to send are counted in failed with a reason each. Up to 100 per call. Each approved reply is posted publicly under the comment. Confirm with the user.

Bulk reject replies. Drops the pending draft for each id. Ids that are not pending are ignored; rejected says how many changed.

List posts. Content Studio posts, newest first, each with its media. status=scheduled is the queue, status=failed needs attention.

Get a post. One post with its media, published ids and any error.

Get posting capabilities. What to read before building a post: which targets are connected, the TikTok creator's allowed privacy levels and video limit (live from TikTok, with a safe fallback), the post types, and the plan's monthly post cap and storage cap with current usage.

Create a post. Creates a draft, or a scheduled post when scheduled_at is set (ISO-8601 with offset; a bare local time is read in the account timezone). Upload media afterwards with POST /posts/{id}/media, then publish or schedule. Rules: TikTok-only posts cannot be a carousel or story and carry no first comment or comment auto-reply; scheduling needs a caption (stories excepted) and a future time; the plan's monthly cap counts scheduled and published posts, a plain draft never holds a slot; a comment auto-reply on a scheduled post needs a free live comment-trigger slot (403 plan_limit_reached). tiktok.privacy_level must be one of the creator's options from GET /posts/capabilities.

Update a post. Edit a draft, scheduled or failed post (409 not_editable otherwise). Send only what changes; tiktok merges field by field; other arrays and objects replace. scheduled_at set moves or schedules it (post.scheduled fires); scheduled_at null turns a scheduled post back into a draft.

Schedule a post. Sets or moves the publish time of a draft, failed or scheduled post. The scheduler publishes it within a minute of that time. Same rules as scheduling on create. post.scheduled fires.

Publish a post now. Publishes to every target at once, the portal's Publish Now. Needs a caption (stories excepted), media for any Instagram target, and two or more items for a carousel (400). A draft, scheduled or failed post can be published (409 not_editable otherwise). When a platform rejects it the post becomes failed and the call is 422 publish_failed with the post and per-platform results. post.published or post.failed fires either way. Send {} as the body.

Unschedule a post. Turns a scheduled post back into a draft. Send {} as the body.

Retry a failed post. Publishes a failed post again, up to the retry cap. Send {} as the body.

Delete a post. Removes the post and its media files. Published posts cannot be deleted (409 not_editable); the live post stays on the platform.

List tracked competitors. Every account you track, with the numbers from its latest stored snapshot, plus how many slots your plan allows. Free 1, Creator 3, Pro 5, Business 10. Whole set in one page. last_snapshot_date is null until the first refresh.

Compare against one competitor. That competitor and your own account side by side, with the per-metric differences the portal shows. Stored snapshots, no Meta call.

Get the leaderboard. Your account and every competitor ranked together: followers, engagement rate, posts per week, average likes and comments, follower growth per day and the content mix (reels, photos, carousels). This is the table behind the portal's Market Intel page. Stored snapshots, so it is instant and spends no Meta quota. Your account and every competitor ranked together. Start here for any "how do I compare" question.

Get follower history. Follower count and engagement rate over the window, one series per account including your own, oldest point first. One point per day that a snapshot was taken, so gaps mean nobody refreshed that day.

Get wins and opportunities. The comparisons already computed against each competitor over the last 30 days: where you are ahead (wins) and where they are (opportunities), each with the metric and the size of the gap. Derived from stored snapshots, no AI involved.

Get audience demographics. The latest stored breakdown of your followers: gender and age as percentages, top countries and cities with counts, and followers online by hour when Instagram supplies it. 404 until the first refresh. Instagram withholds all of it for accounts under roughly 100 followers, which comes back as zeros rather than an error. Includes followers online by hour, which is what "when should I post" is answered from.

List your top posts. Your own recent posts with their stored performance: reach, impressions, likes, comments, saves, shares, video views, engagement rate and save rate. Newest first. Stored, so refresh first for current numbers.

Get account insights. Reach, profile views, interactions, saves, shares and follows for your own account over the last 1, 7 or 30 days. This one calls Instagram live, so it takes a moment and counts against Meta's rate limit. 403 permission_required when the account was connected without the insights permission; reconnect it in the portal to fix that.

Track a competitor. Looks the handle up through Meta and starts tracking it. The account must be a public Instagram business or creator account; personal and private accounts cannot be read by anyone through the API (422 invalid_param). At the plan cap this is 403 plan_limit_reached, and an account you already track is 409 conflict. The first snapshot is taken immediately.

Stop tracking a competitor. Frees the slot. History is kept, so re-adding the same handle picks up where it left off.

Refresh from Meta. Pulls current numbers from Meta for your account and every tracked competitor and writes today's snapshot, exactly like the portal's Refresh button. One Meta call per account, so it takes several seconds; every other Market Intel read serves the snapshots this writes. Five minutes are enforced between refreshes (429 too_soon carries retry_after_seconds). Send {} as the body. Slow: one Meta call per tracked account. Only call it when the stored snapshot is stale.

List FAQs. Question and answer pairs the AI answers from. Active only by default.

Create an FAQ. Counts against the plan cap. The answer is embedded for retrieval right away; embedding_pending is true if indexing is still catching up.

Update an FAQ. Partial update. Changing question, answer or keywords re-embeds it; changing only active does not.

List services. Services and pricing the AI can quote and, with booking on, offer. Active only by default.

List knowledge gaps. Questions the AI could not answer well, flagged by the gap detector with the customer message, the AI reply, a canonical form of the question, and a suggested answer and target. Open gaps by default; status=all for everything. Read-only: fix a gap with POST /knowledge/faqs or by editing identity.instructions, and dismiss it in the portal.

Get knowledge base usage. How many items exist against the plan cap, broken down by FAQs, services and documents, plus document storage. Check can_add before creating.

Get an FAQ. One FAQ by id.

Delete an FAQ. Removes the FAQ and its embeddings.

Get a service. One service by id.

Create a service. Counts against the plan cap. is_virtual is honoured only when Google Calendar is connected and Meet links are on; otherwise it is stored as false. Images are added in the portal.

Update a service. Partial update. Any content change re-embeds; changing only active does not. The service image is never touched by the API.

Delete a service. Removes the service, its embeddings and its image file.

List documents. Uploaded PDFs and text files. Content is not returned, only metadata and embedding state. Upload happens in the portal.

Get a document. One document by id, metadata only.

Rename or toggle a document. Change the name or switch it off without deleting it.

Delete a document. Removes the document and its embeddings.

Get the AI configuration. All nine sections in one call. available on file_handling, followups, csat and comment_ai says whether the plan includes that feature.

Get the per-channel AI switches. For meta (Facebook + Instagram share one switch), sms, whatsapp and telegram: whether the channel is connected, whether the AI answers on it, and whether automations run on it.