PappaChat

Developers: connect your AI

The panel is the visual way to run the account. There is a second one: connecting your own AI (any client that supports the Model Context Protocol, MCP) and running the account in your own words. "Create an assistant for the sales mailbox", "what bookings do I have this week", "add the new opening hours to the Knowledge base": your AI does it, inside the permissions you gave it.

Everything starts from the Developers section of the panel, and it is reserved for the Max plan.

Two ways to connect

  • Guided authorisation. From your AI client you add PappaChat as a connector: the browser opens, you sign in with your account and choose what to authorise. This is the road for anyone using a ready-made AI app.
  • API key. From the Developers section you generate a key with the same permissions chosen at creation, bound to your organisation. This is the road for your own scripts and integrations.

Either way, the connection is revoked from the panel in one click, and every call is recorded in a log you can read.

You choose what it may touch

The authorisation is not a yes or no: it is a list of areas, and for each one you choose between read and manage. You can give an AI nothing but read access to analytics. A connection can never widen itself: on renewal it stays what it was, and to give it more you have to authorise again yourself.

The areas, and what they mean in practice:

  • Assistants. Read: the list of your assistants. Manage: create them, change their personality and behaviour rules, assign them to a channel, switch the bot on and off on a channel.
  • Knowledge base. Read: see what the assistant knows. Manage: add content, set the automatic refresh from your website, close the questions the assistant could not answer.
  • Reservations. Read: availability, bookings, settings, the approval queue. Manage: create, move and cancel bookings, block out time, manage working hours, resources and the approval policy.
  • Conversations. Read only: the list and the messages. Replying to a customer is not available, by choice: the moment an external AI can write to a real person deserves a permission of its own, which today does not exist.
  • Contacts and leads. Read: the captured contacts and the forwarding contacts. Manage: update a lead's status, manage forwarding contacts.
  • Channels. Read: connected channels, message templates, numbers, the WhatsApp profile. Manage: connect Telegram, manage message templates, the email identity, number settings and the WhatsApp profile.
  • Integrations. Read: query the connected business system (products, stock, order status), always inside the permissions you gave the integration itself. Manage: connect, configure and disconnect integrations.
  • Analytics. Read only: channel numbers and reports.
  • Account and settings. Read: account status, settings, pending items. Manage: change settings, apply preset configurations, answer a consultation from the assistant during a live call.
  • Support. Open and read your requests to our support.

The second gate: the person

A permission given to the connection never outranks the person who created it. Every call is worth the intersection of what the connection may ask for and what that person may do in the panel: an operator with limited access gains nothing by connecting an AI, and if their permissions change, they change for the connections already open too. Connecting at all requires a dedicated permission, which the administrator grants and removes like any other.

What it cannot do, by construction

  • It does not write to your customers. No outbound message ever reaches a real person.
  • It does not delete destructively. Irreversible operations are not exposed.
  • It does not leave your organisation. Every call is bound to your account, and another business's data does not exist, as far as it is concerned.
  • It does not work in the dark. Every call lands in the Developers section's log, with who, what and when.