It works toward an objective
A chatbot replies. An agent has a finish line — quote sent, date confirmed, invoice paid — and moves the conversation column by column until it gets there or hands over.
WhatsApp AI agent
Most WhatsApp AI is a chatbot with a new name. A Calypso Agent is bound to one board, runs on that board’s knowledge, and works every conversation toward the board’s objective — under boundaries you set.
A chatbot replies. An agent has a finish line — quote sent, date confirmed, invoice paid — and moves the conversation column by column until it gets there or hands over.
Bookings, orders, follow-ups, CRM writes. The agent performs the actions you permit; sensitive or irreversible ones wait for a person to approve them.
Every answer, action, and move is logged with its source and the person or agent behind it. You can read back any conversation and see exactly what happened and why.
01
The agent is bound to a single board, so its knowledge and its behaviour are scoped to that job. Support does not answer with the sales price list.
02
A Hidden board it may never read, guardrails on what it may quote or confirm, and the list of actions it is allowed to perform on its own.
03
Pause the agent and your team is on the board instantly, mid-conversation, with full context. Flip it back and it resumes where it stopped.
A chatbot answers a message and waits for the next one. An agent is given an objective and the means to reach it: it can look things up in your knowledge, take permitted actions like a booking or a CRM update, and move a conversation through stages until the objective is met or a person takes over.
Only within the permissions you set. Routine actions follow your rules. Anything sensitive or irreversible waits for approval, and the Hidden board marks contacts and chats it may never read at all.
That is your call, and we recommend saying so. The agent can introduce itself as your assistant, and it always hands over to a person when the conversation calls for one.
Yes — one per board, each scoped to its own department, knowledge, and behaviour. Plans run from a single agent up to unlimited.
Yes. The same agent is reachable through an OpenAI-compatible API, MCP, an n8n node, and an embeddable web widget, running on the same knowledge as the board.
Your knowledge stays isolated to your workspace, is never shared with other customers, and is never used to train shared models. You can export or delete it.
Start today
Bound to its board, in your voice, working toward your objective — and one switch from off.