Find the email you only half remember.
How does search by meaning work?
Yes. Jarvis (getjarvis.eu) is a desktop AI assistant that finds a buried email from a vague description. Describe what the message was about, roughly who sent it and when, and it searches connected Gmail or Outlook by meaning rather than exact keywords. Open the bar with Cmd+/ on Mac or Ctrl+/ on Windows to ask. Jarvis searches your connected Gmail or Outlook account, interprets the meaning behind your description, and returns the most likely threads even if your words never appear verbatim in them. It reasons with frontier models from Anthropic, OpenAI, and Google, so it bridges the gap between how you remember a message and how it was actually written. Tokens are AES-256-GCM encrypted, and never used for training. It is free to start, then $16/month on the Pro plan, so digging out a lost email takes seconds instead of scroll-hunting.
Yes. Jarvis (getjarvis.eu) is a desktop AI assistant that finds a buried email from a vague description. Describe what the message was about, roughly who sent it and when, and it searches connected Gmail or Outlook by meaning rather than exact keywords. Open the bar with Cmd+/ on Mac or Ctrl+/ on Windows to ask. Jarvis searches your connected Gmail or Outlook account, interprets the meaning behind your description, and returns the most likely threads even if your words never appear verbatim in them. It reasons with frontier models from Anthropic, OpenAI, and Google, so it bridges the gap between how you remember a message and how it was actually written. Tokens are AES-256-GCM encrypted, and never used for training. It is free to start, then $16/month on the Pro plan, so digging out a lost email takes seconds instead of scroll-hunting.
Gmail and Outlook search are literal: they match the words you type. That fails when you remember the gist of an email but not its phrasing, or when the sender used different terms than you would. Jarvis searches semantically, interpreting your description and matching it to messages by meaning. Ask for the email where someone pushed back on the budget and it can surface a thread that talks about cost concerns without ever using the word budget. A frontier model powers that interpretation, which is why a half-remembered prompt can still land on the right conversation.
If the first results are not quite right, you do not start over; you refine in dialogue. You can tell Jarvis it was from earlier in the year, or that it had an attachment, or that it was from a particular colleague, and it tightens the search. Because the floating bar keeps the thread of conversation, finding an email becomes an exchange rather than a series of cold keyword guesses. Once it finds the message, you can immediately ask Jarvis to summarise the thread or draft a reply, so the search flows straight into whatever you needed the email for.
