Autocalls AI

Create a folder to organise assistants.

create-folder

Create a new knowledgebase the assistant can search during calls. Add documents to it with create-document.

create-knowledgebase

Create a label to tag and organise assistants.

create-label

Add a lead (contact) to a campaign the user owns. Only add people who agreed to be contacted. The phone number is validated and normalised to E.164. Optionally pass variables to personalise the call and secondary_contacts to chain fallback numbers.

create-lead

Create a reusable custom mid-call tool the assistant can call during a conversation: an HTTP endpoint plus the parameters it should collect. Use {param} (single brace) in the endpoint/body for assistant-collected parameters, and {{system variables}} (double brace — {{customer_phone}}, {{assistant_phone}}, {{assistant_id}}, {{assistant_name}}, {{current_date}}, {{current_time}}) in the endpoint, header values or static fields; those are substituted automatically at call time. This creates an HTTP-type tool. Automation-Platform tools (which auto-build a connected flow with no manual webhook setup) are created in the app UI, not here.

create-mid-call-tool

White-label owners only: create an access token for one of your OWN platform's end-users (by user_id or email), so that user can call this platform's REST API. This is the end-user's own platform credential, not a third-party secret. Returns it once — show it to the owner once so they can pass it on securely, since it cannot be shown again.

create-platform-user-token

Revoke an API key. Any integration using it will stop working. This cannot be undone.

delete-api-key

Permanently delete an assistant the user owns. This cannot be undone.

delete-assistant

Permanently delete an automation the user owns. This cannot be undone, and everything the automation did stops. If you are deleting it to REPLACE it and the user still wants any of its current behavior, call get-automation FIRST and carry those steps into the replacement — they are unrecoverable after deletion. Use it to clean up automations that failed their test run or are no longer needed.

delete-automation

Permanently delete a call record the user owns. This cannot be undone.

delete-call

Permanently delete a campaign the user owns. If it is currently running it will be stopped first. This cannot be undone.

delete-campaign

Permanently delete a document from a knowledgebase the user owns. This cannot be undone.

delete-document

Delete a folder the user owns. Assistants inside it are not deleted, just un-foldered.

delete-folder

Permanently delete a knowledgebase the user owns, including its documents. This cannot be undone.

delete-knowledgebase

Delete a label the user owns. It is removed from any assistants tagged with it.

delete-label

Permanently delete a lead the user owns. This cannot be undone.

delete-lead

Permanently delete a custom mid-call tool the user owns. It will be detached from any assistants currently using it. This cannot be undone.

delete-mid-call-tool

Delete a BYO SIP trunk the user owns. (SIP trunks are stored as phone numbers, so this releases the underlying number via the phone-numbers endpoint.) This cannot be undone.

delete-sip-trunk

Pause AI auto-replies on a conversation (identified by its uuid from list-conversations), e.g. to let a human take over.

disable-conversation-ai

Re-enable AI auto-replies on a conversation (identified by its uuid from list-conversations), so the assistant resumes answering automatically.

enable-conversation-ai

Generate a single AI reply for a message using one of the user's assistants (its prompt, knowledgebase and tools). Costs account balance. Intended for one-off testing: it reuses or creates a conversation record behind the scenes to hold context — to manage an ongoing chat yourself, use create-conversation + send-conversation-message instead.

generate-ai-reply

Get the authenticated user's account information, including name, email and current balance. Useful before actions that cost money (calls, SMS, WhatsApp, buying numbers).

get-me

Get the full configuration of a single assistant the user owns: system prompt, initial message, voice, model, language, behaviour settings, built-in tools and attached mid-call tools. Calendar credentials are masked for security — when updating tools, omit those fields (or leave them blank) and the existing ones are preserved. Use this to inspect an assistant before editing or improving it.

get-assistant

Get a single automation the user owns, by id (from list-automations). Returns its full definition so you can inspect the trigger and its steps.

get-automation

Get the full detail of a single call the user owns, including the complete transcript, extracted variables, post-call AI evaluation, costs and recording URL. Use this to diagnose how an assistant performed on a specific conversation before improving it.

get-call

Get a single chat conversation by its uuid, including its messages. The response also carries the conversation's assistant as an object {id, uuid, name} — use that integer assistant.id with get-assistant.

get-conversation

Get a single document from a knowledgebase the user owns (status, type, source).

get-document

Returns the platform's authoritative how-to guides. Useful before building, fixing or configuring an assistant; writing a system prompt; setting up a mid-call tool, inbound recognition or CRM lookup; sending an SMS or WhatsApp during a call; getting or connecting a phone number, SIP trunk or caller ID; or answering an account, billing or setup question. Call with no topic to list the available guides and what each covers, then call again with a topic key to get the full guide for that topic.

get-guide

Get a single knowledgebase the user owns, including its documents and status.

get-knowledgebase

Get the full definition of a single custom mid-call tool the user owns (type, endpoint, method, body format, headers and static fields with their secret values masked, and parameter schema). For an automation-type tool the endpoint is a managed flow URL; manage that tool in the app UI.

get-mid-call-tool

Get the configuration of a single SIP trunk the user owns.

get-sip-trunk

Get an example payload for an assistant webhook, to help the user build their automation. webhook_type: post_call (after a call) and conversation_ended (when a chat ends) — these match configure-assistant-webhook; plus conversation (a chat/conversation event) and whatsapp_voice (Ask to call / permission / WhatsApp call started). NOTE: there is no sample for the "inbound" pre-call-variables webhook (configure-assistant-webhook type "inbound") — that one just returns the flat key-value variables you define on the assistant.

get-webhook-sample

List the account's API keys (names and last-used dates only - never the secret token).

list-api-keys

List the AI voice assistants owned by the authenticated account. Returns an object with a `data` array of assistant summaries (id, name, type, mode, folder, labels) plus `page`, `total_pages` and `has_more`; read from `data` and call again with the next `page` while `has_more` is true. Call get-assistant for the full configuration of a single assistant.

list-assistants

Resolve the REAL choices for an integration step's dropdown field (trigger OR action) using the user's own connected account — e.g. their Facebook pages/lead forms, CRM pipelines, spreadsheets. Call this BEFORE building an automation on an external integration and let the user pick from the returned list; NEVER guess ids. Use the returned value, not the label. Requires the integration's account to be connected already.

list-automation-options

List an automation's recent executions (newest first), or — when run_id is given — one execution's step-by-step detail with a per-step classification (wiring_error = fix the definition; missing_connection / missing_record = expected, not a definition bug). Use this to debug why an automation is failing. To re-run a failed execution, use retry-automation-run.

list-automation-runs

Ready-made automation templates. ALWAYS check this list FIRST when the user asks for an automation — if a template matches, use apply-automation-template with its params instead of writing a definition from scratch. Each entry lists its params, required account connections, whether it attaches to an assistant, and whether it sends real messages (side effects).

list-automation-templates

List the automations in the user's account. Returns an object with a `data` array of automation summaries (id, name, status). Use get-automation for a single automation's full definition.

list-automations

List the calls belonging to the user's assistants, most recent first. Returns an object with a `data` array of compact call records (no transcript) plus `page`, `total_pages` and `has_more`; read the records from `data`, and to get older calls call again with the next `page` while `has_more` is true. Use get-call on a `data[].id` for the full transcript and evaluation. Supports filters. Ideal first step when analysing or improving an assistant.

list-calls

List the user's campaigns (call, WhatsApp, SMS): id, name, channel, status, schedule_windows, messages_per_minute, WhatsApp/SMS config ids, and fallback settings.

list-campaigns