Pranan started in your inbox. It reads your sent mail, learns how you write, keeps track of who matters and what you promised, and leaves drafts in Gmail for you to send. That knowledge is your Twin.
The trouble is that a lot of work no longer starts in the inbox. It starts in a chat with an AI assistant, and that assistant has no idea who Anika at Northwind is, what you told her last week, or that you owe her a revised quote by Friday. So you paste in context, or you go without it.
The Pranan connector closes that gap. It is in early access, in Claude and in ChatGPT, and it works with other apps that support MCP connectors.
What you can ask
Once Pranan is connected, you ask in plain words and the assistant answers from your Twin.
- Who is this? "Who is Anika at Northwind and where do things stand?" You get their role, the state of the relationship, recent memory, open commitments, the latest emails and upcoming meetings.
- Prep my next meeting. "Prep me for my 3pm." Everyone in the meeting, what you owe each of them, and the latest threads, before you walk in.
- What do I owe people? The promises you made in email, overdue first, each linked to the email you made it in.
- Search what you know. "What did we agree on Northwind pricing?" Memory and email together, newest first, with sources.
- Write like you. Your role, your writing style and the writing rules you kept, so the assistant drafts in your voice.
There is one optional extra. If you allow it when you connect, you can ask the assistant to remember a note about a person. The note is saved privately to your Pranan memory and used in your drafts. That box is off by default.
What it cannot do
The connector reads. It cannot send, draft or delete email, and it cannot change anything in your mailbox. Pranan does not request permission from Google to send mail at all. That is the same line Pranan has always drawn: you send your own email.
A few other limits are built in:
- No app reads anything until you press Allow in Pranan, and each connected app is listed in your settings with when it was last used.
- Answers carry short excerpts and a link to the email in your own Gmail, not whole messages.
- A teammate's private memory is never returned. Personal facts you marked sensitive never leave Pranan. People who asked to be erased are never named.
- Every request is logged by which tool ran and when, never by what it returned.
One thing to know before you connect: once your AI app receives an answer, it handles that answer under its own terms and your settings there. Our privacy policy covers what Pranan shares and when.
How to connect
- In Claude, add a custom connector with the address
https://app.pranan.ai/api/mcp. In ChatGPT, or another app that supports MCP connectors, add the same address. - Sign in to Pranan. You see exactly what the app will be able to read, and you press Allow or Cancel.
- Ask. The assistant picks the right Pranan tool and cites the email or memory each answer came from.
To stop, open Pranan settings, go to Connected AI apps and press Disconnect. Access ends on the next request.
Why we built it this way
An assistant that knows the internet but not your relationships gives you generic help. An assistant that can read everything and act on its own is a risk most people should not take with their inbox. The connector sits between the two: your AI app gets the context it was missing, with sources, and the decisions stay with you.
You can start on the Free plan. Connect Gmail first so your Twin has something to learn from, then add Pranan in your AI app. The full details are on the connector page.