Answer in Pachca conversations as your own bot

The agent works through the Pachca Bot API at api.pachca.com and passes the bot token in the Authorization header as a Bearer token. Its first request is GET /profile: that checks the token is alive and shows whose name the messages will go out under. From there only the conversations the bot has been added to are in play — their threads, attachments and button taps.

A bot in Pachca is a conversation participant, not an observer on the side: until it is added to a conversation the agent receives not a single message from there, and no permission gets around that. Incoming messages arrive not by polling but over an outgoing webhook: the address is registered at connection time, Pachca pushes new messages and button taps itself, and signs every request with an HMAC in the Pachca-Signature header.

Writing is the main mode for this connector. The agent answers in the thread, uploads files, asks follow-up questions with buttons and handles the tap by its trigger_id. While it thinks, a reaction sits on the original message — Pachca has no typing indicator, and the reaction honestly shows the request was picked up. It can edit and delete only its own messages.

The token carries exactly this bot's rights and nothing beyond: the profile owner's personal correspondence is out of reach through it. Attachments are fetched by re-reading the message by its id, and the agent holds the request rate itself — Pachca meters reads separately, and the agent does not spend an extra call on every message. A file goes out as an upload to Pachca rather than a link to an outside drive, and the agent names the conversation it landed in.