Wenlan
Wenlan is a source-backed AI knowledge base and LLM wiki for AI work: organize documents and decisions into pages you can find, inspect, and review.
翻訳待ち
Calendar, mail, bills, lists and reminders, on your own computer, with a privacy guard that keeps card and account numbers away from the AI.
翻訳待ち
生産性TypeScript · JavaScriptGitHub @Ruihang2017オンライン所有者未確認
Have you tried it? One tap is enough. No sign-in needed.
Loading the full README. The saved excerpt is shown below.
A personal assistant for your terminal that does things: keeps your calendar, reads and drafts email, tracks bills, reminds you — and removes account numbers, card numbers and passwords before anything reaches the AI. Edward runs on your own Windows computer. It uses your own ChatGPT account through the Codex CLI and, if you connect them, your own Google Calendar and Gmail. There is no Edward server. Status: developer preview. Windows only. You bring your own ChatGPT account, Codex CLI and (for calendar and mail) Google Cloud project. Expect rough edges. [README 설치·사용 부분 발췌] InstallRead the full README ↗
A personal assistant for your terminal that does things: keeps your calendar, reads and drafts email, tracks bills, reminds you — and removes account numbers, card numbers and passwords before anything reaches the AI.
Edward runs on your own computer, Windows or Mac. It uses your own ChatGPT account, or your own OpenAI API key, through the Codex CLI and, if you connect them, your own Google Calendar and Gmail. There is no Edward server.
Status: developer preview. Windows, and macOS (Intel and Apple silicon). You bring your own ChatGPT account or OpenAI API key and (for calendar and mail) Google Cloud project; the terminal version also needs the Codex CLI. Expect rough edges.
▶ Watch the two-minute demo with a voice-over and captions (the picture above is the same video, four times faster). The website has a short video for each part. Everything shown is made-up demo data.
In the terminal it looks like this:
you › what's on tomorrow, and is anything due this week?
📅 checked calendar 2026-10-03 → 2026-10-03
💳 checked bills
edward › Tomorrow: dentist at 15:00. Your electricity bill ($245.30) is due Sunday.
you › reply to Sam's update and ask whether the energy account is set up
✉ searched mail "Weekly update"
✉ read conversation m2
✉ drafted reply to m2
⚠ Allow gmail_send?
send email to Sam Taylor: RE: Weekly update
[y] yes [n] no
| Chat | Answers, web search, image generation, with memory of what you tell it |
| Reminders | "Remind me at 5pm to call the dentist" — a notification, even when Edward is closed |
| Calendar | What's on, when you're free, add / move / delete events (each one confirmed) — across all your Google accounts |
| Gmail | Search, read, summarise; draft replies; send only after you approve the draft — one or several personal Gmail accounts |
| Lists | Shopping list, things to fix at home, to-dos, in Google Tasks so they are on your phone too |
| Voice (desktop app) | Talk to Edward and hear the reply, with your own OpenAI API key |
| Bills | Finds bills in your email, checks them for signs of fraud, reminds you before they are due |
| Daily brief | Today's events, reminders, bills and unread mail in one glance |
| Heads-ups | Before an event with a place or people (and the related emails); a mail summary twice a day; the week ahead on Sunday evening; trips from booking emails, with when to leave |
Three ways in (the desktop app, the terminal, a background job that runs for a moment every minute) share one core. The core reaches the AI only through Codex, a process on your own computer, and everything on that path passes the privacy guard. Calendar, mail and lists come from Google with your own sign-in. There is no Edward server.
The model only reads and writes text. It can't send an email or change your calendar: it asks for one of Edward's tools, Edward's own code does the work, and anything that changes something waits for your OK. What comes back from outside (an email, a calendar entry) is cleaned and passes the guard before the model sees it. The model itself is your choice: a ChatGPT plan, an OpenAI API key, or another service that speaks OpenAI's Responses format, such as OpenRouter or Qwen on Alibaba Cloud (experimental).
Before anything is sent to the model, Edward's own code removes card numbers, bank and account numbers, passwords, PINs, one-time codes and ID numbers — partial ones too ("card ending 1234"). They are never stored. This is a filter at the process boundary, not an instruction to the model.
Edward cannot pay anything or sign in to a bank. Sending an email, changing your calendar and marking a bill paid all need your approval.
What it does not catch, and the rest of the security model, is in SECURITY.md. Read it before you rely on Edward.
This is the beginning of the README. Read the rest at the original. Read the full README ↗