Tools

The 22 MCP tools organized by category. ✅ marks required parameters.

Assistants

Tool	Description	Main parameters
`list_assistants`	Lists all assistants for the account	`page`, `perPage`, `name`
`get_assistant`	Returns full configuration of an assistant	`id` ✅
`create_assistant`	Creates a new assistant	`name` ✅, `prompt`, `language`, `temperature`, `maxTokens`, `restrictedByContext`
`update_assistant`	Updates fields of an existing assistant (only the provided fields are changed)	`id` ✅, `name`, `prompt`, `isActive`, `language`, `temperature`, `maxTokens`, `restrictedByContext`

Valid values for language: brazil-portuguese · english · any

temperature: between 0.0 and 2.0 (recommended 0.3–0.8 for customer-facing bots). maxTokens: maximum tokens per response: 0 or omitted uses the model default. restrictedByContext: when true, the assistant only answers questions grounded in the uploaded knowledge base. isActive: deactivates the assistant without deleting any configuration.

Triggers

Tool	Description	Main parameters
`list_triggers`	Lists triggers for an assistant	`assistantId` ✅, `page`, `perPage`
`get_trigger`	Returns full definition of a trigger	`assistantId` ✅, `triggerId` ✅
`create_trigger`	Creates a function calling trigger	`assistantId` ✅, `name` ✅, `description` ✅, `parametersJson`, `isActive`, `interpretResponse`, `copilotOnly`
`update_trigger`	Updates an existing trigger (auto-merges with current state)	`assistantId` ✅, `triggerId` ✅, `name`, `description`, `parametersJson`, `isActive`, `interpretResponse`, `copilotOnly`
`delete_trigger`	Removes trigger and all its actions (irreversible)	`assistantId` ✅, `triggerId` ✅

name must be camelCase (e.g. searchOrder). description is max 350 chars and must explain exactly when the LLM should call the function.

parametersJson is a JSON Schema serialized as a string describing the arguments the LLM must collect before invoking the trigger. Supported types: string, number, boolean, array (with items). Every property must include a description (max 200 chars).

interpretResponse: true lets the LLM rephrase the action result in natural language; false forwards the response as-is. copilotOnly: true makes the trigger available only in copilot (agent-assist) mode.

Actions

Tool	Description	Main parameters
`list_actions`	Lists actions for a trigger	`assistantId` ✅, `triggerId` ✅, `page`, `perPage`
`get_action`	Returns an action's configuration	`assistantId` ✅, `triggerId` ✅, `actionId` ✅
`create_action`	Creates an action inside a trigger	`assistantId` ✅, `triggerId` ✅, `type` ✅, `actionJson`
`update_action`	Replaces an action's configuration	`assistantId` ✅, `triggerId` ✅, `actionId` ✅, `type` ✅, `actionJson`
`delete_action`	Removes an action from a trigger (irreversible)	`assistantId` ✅, `triggerId` ✅, `actionId` ✅

Available action types (type):

Type	Description
`WEBHOOK`	Custom HTTP request (GET/POST/PUT/DELETE with headers and body)
`NATIVE`	Native WhatsApp messages: text, buttons, list, image, audio, file, AI-generated audio, assistant transfer
`CRM`	Upsert a contact in Verboo's built-in CRM
`GOOGLE_CALENDAR`	Check availability, list events, schedule meetings
`META_ADS`	Conversion event for the Meta Pixel (Purchase, LeadSubmitted, etc.)
`FEEGOW`	Patients, appointments, professionals, specialties, insurance
`CHAT_INTEGRATION`	Change conversation status or summarize it in the panel
`KNOWLEDGE`	Semantic search across the assistant's knowledge base
`Z_API`	Z-API WhatsApp: apply tag, reset session
`MY_ZAP`	MyZap WhatsApp: tag, reset, send file/image, self-message
`EVOLUTION`	Evolution API WhatsApp: tag, reset, send file/image, mark unread
`CATALOG_INTEGRATION`	Search products, manage cart (search, get/add/remove cart)

actionJson is a JSON string with the type-specific configuration. To reference a trigger parameter inside an action field, use the parameter name as a plain string (no curly braces, no {{ }}). The runtime substitutes it with the value the LLM collected. Exceptions: WEBHOOK body and NATIVE texts/baseText are literal content.

Sessions (test agents end-to-end)

Tool	Description	Main parameters
`create_session`	Creates a new conversation session with an assistant. Returns the `sessionId` for follow-up calls	`assistant_id` ✅
`get_session`	Retrieves the full state of a session, including all messages exchanged	`session` ✅
`send_message`	Sends a user message to a session and returns the assistant's response	`session` ✅, `message` ✅
`delete_message`	Removes a specific message from the history so the session can be re-tested with a different path	`session` ✅, `message_index` ✅

Use these tools to validate prompts, triggers, actions, and the knowledge base without connecting a real WhatsApp number. Typical flow: create_session → send_message (multiple times to simulate the user) → get_session to inspect the full transcript.

History & Analytics

Tool	Description	Main parameters
`list_generative_messages`	Lists individual conversation messages (content, role, tokens, timestamps)	`assistantId`, `sessionId`, `page`, `perPage`
`list_generative_sessions`	Lists unique sessions with message counts and timestamps	`assistantId`, `page`, `perPage`
`get_generative_message_graph`	Daily message-volume time series	`assistantId`, `range` (1–365 days, default 7)
`list_assistants_usage`	Messages per assistant for the current month, sorted high-to-low	:

Want to see all this in action? Check out 3 real conversations → showing how Claude chains these tools end-to-end.