Decisions, promises and numbers from Slack, Telegram and email, kept in one place. Ask about them in any chat, or read a two-minute digest every morning.
Slack channels and Telegram groups. Penny also reads the history you choose.
Small talk stays in the chat. What matters goes to memory, with a link back to the original message.
Answers come with sources. The digest arrives in Telegram, Slack or email.
Chats, email, your database and your code. Penny reads them all the time and keeps one memory of what was decided, promised and said, and by whom. Then you use it wherever you are.
Same memory, different questions. Choose a role to see a real kind of question, where the answer comes from, and what it replaces.
You're in twelve chats and three inboxes. Decisions happen in threads you skim at midnight. Penny reads all of it, so the week's picture is one question away.
Terms get agreed on a call, confirmed in email and tweaked in a shared Telegram chat. Penny keeps all three together, so nobody quotes a different price twice.
The code says what. The reasons live in an incident thread, a PR review and an open issue. Connect Penny to Claude Code or Cursor, and the agent checks them before it touches anything.
The database knows how many customers left. Support threads know why. Penny reads both, so you get the count and the story in one answer, without asking an engineer for SQL.
Every portfolio company has its own Telegram group, and every one of them asks you for something. Add Penny to the groups, and nothing you promised gets lost in the scroll.
Examples. The companies, people and numbers are made up; the kinds of questions and sources are real.
Simple mode is a digest, a question box, your sources and your balance. Advanced mode is every fact, person and thread, with search and history. Same data, one switch.
Thursday, 25 September. I read 38 conversations while you slept.
What I read, and what you can add. Everything connected is seen by the whole team.
The same console on your phone
No new app to open. Mention @Penny in a Slack channel or a Telegram group, and she replies in the thread with links to the exact messages. Everyone in the chat sees the answer, so the question gets asked once.
Telegram group, on the phone
Penny speaks MCP, the standard way AI apps connect to tools. Add one address to Claude, ChatGPT, Cursor or Claude Code, and your agent can search the memory, read the sources and ask Penny, with the same citations you get.
https://app.ipenny.app/mcpCopyhttps://app.ipenny.app/mcp/askCopyclaude mcp add --transport http penny https://app.ipenny.app/mcpThe client opens a Penny page in your browser: sign in, check what the agent asks for, click Allow. No keys to copy.
Most answers need both: a number from the database and the thread that explains it, a line of code and the decision behind it. Penny reads all of it, so you don't have to stitch it together.
Connect Postgres with a read-only user you create, and choose the tables Penny may read. You and your agents get live numbers, with the team's context around them. Penny never writes to your database.
order_id | status | shipped_at | refund |
|---|---|---|---|
| 10492 | shipped | 2026-09-23 | null |
Install the GitHub app and pick the repositories. Penny reads the code, pull requests, issues and commits, and links them to the conversations that led to them. Your coding agents can search and read the code through Penny too.
web/config/billing.ts
- export const TRIAL_DAYS = 14;
+ export const TRIAL_DAYS = 7;Penny sends what changed since yesterday: new decisions, promises made, what's overdue and the numbers that moved. In Telegram, Slack or email, wherever you read first.
Penny reads a lot, so we keep the rules short and strict.
Only the channels, chats, inboxes, repositories and tables you connect. Databases are read-only, and you can pause any source.
We call AI models with zero data retention. The model provider doesn't keep your messages after answering.
Not by us, not by the model providers. Your conversations teach nobody's model.
Anyone in a Telegram group can send /forgetme. Remove a source and what Penny learned from it goes too.
You choose every source, channel and table. Penny tells a group when she joins, so nobody is read by surprise.
Invite teammates with a link. Everyone in the team sees the same memory; nobody outside it sees anything.
$10free credit on signup
After the credit runs out you pay what the AI costs, plus 50%. No subscription and no per-seat fee. Top up from $10. If the balance hits zero, Penny queues new messages and catches up after a top-up.
Start with $10 free creditNo. She reads only the channels, groups, mailboxes, repositories and tables you connect. You can add or remove any of them at any time.
Only when someone mentions @Penny or messages her directly, and with the morning digest in the chat you choose. She doesn't chime in otherwise.
Everyone in the team sees everything Penny knows. That keeps it simple: connect the channels your team already shares, and keep private ones out.
Yes. Penny is an MCP server: add app.ipenny.app/mcp to Claude, ChatGPT, Cursor, Claude Code or any MCP client, then sign in and allow access.
Never. Models are called with zero data retention, so providers keep nothing. You can delete a source, and what Penny learned from it, right away.
The $10 credit covers a year of Slack history for a small team and about three weeks of daily use after that. Then you pay the AI cost plus 50%, with no subscription. Penny shows the price before any big job.
Yes. Make one team per client, so each client's memory stays separate, and switch between teams in the console.
A few minutes to connect Slack or add the Telegram bot. Penny starts on new messages right away and reads the history you choose in the background. Your first digest is ready within minutes.
Connect one chat today. Ask your first question in five minutes. The first $10 is on us.