What's yours stays yours.

Nothing to replace

Your CRM, mailbox and tools stay where they are. Celorus works inside the AI tools your team already uses.

Never pooled with another firm's

Your desk is never pooled with another firm's. When the record can't answer a lookup, we use that to improve it, and what you looked up is never shown to another firm.

Every fact carries its source

Every fact arrives with its source, down to the document and the page.

If the record does not say it, Celorus says “not available” instead of guessing.

Personal data, handled with care

Personal data moves under the discipline India's data protection law asks for, by default.

People named in the record can read what we hold and how to reach us on the notice page. How this website handles your details is on the privacy page.

Where it runs

In the AI tools your team already uses. The ones we support today are Claude, Codex and Kimi.

It runs where your people already spend their day.

What reaches our server

• Your sign-in, by Google or by email.
• When your AI asks for a page: the desk pages it needs, as text, and the words your AI wrote for it.
• Lines from your mail, calendar or CRM, when your AI copies them onto your desk pages or hands them to a desk change.
• The names of files a page points to. Never what is in them, and not kept.
• On Solo, your desk folder in Drive: read when your first desk is made, every day at 06:00 IST to write the day’s board, and when you ask for the latest.
• On Solo, the key Google gives us to reach your desk, and your folder’s id, stored encrypted.
• A usage line for every call: the tool, the time, the size, your account and your firm.
• A company or person name you looked up that the record does not hold.
• The questions your AI asks the record, as text.
• When your AI asks for a research pack: the lead’s name, the name of the lead’s page on your desk, your note, and the country and language to search in, as text. Our server refuses a note in which it finds a phone number, an email or an identity number.

What never reaches our server

• Any file in your Drive that our server did not make.
• Your mailbox, calendar and CRM themselves. We never connect to them. Only what your AI copies onto your desk reaches us.
• Your AI’s own web searches. They run on your own subscription.
• Pictures and other files that are not text, hidden folders, and files no page names.
• Your colleagues’ pages, when you make one of your own.

Who else touches your data

• Your own AI tool, Claude, Codex or Kimi, on your own account.
• Amazon Web Services: our servers, database, keys, logs and sign-in.
• Google: sign-in, if you choose it, and on Solo, your Drive.
• Anthropic’s Claude Haiku, on our own account: when your AI asks the record a question, it reads the question’s text to choose which sections answer it.
• Google’s Gemini, on Google Cloud’s Vertex AI: when your AI asks for a research pack, it reads the lead’s name, your note and the text of the web pages our own systems read for it. These calls may run outside India.
• Brave’s web search, on our own account: for a research pack, it gets the searches Gemini writes, with the country and language to search in. A search can carry the lead’s name and words from your note. Brave keeps searches in its logs for up to 90 days.
• Paddle and Razorpay, for payments only.
• This website: Cloudflare serves the pages and counts visits without a cookie. Google serves the fonts, so it sees the address your visit comes from. The working-session form and the manifesto sign-up go to a sheet in our Google Workspace, with an alert mail to our inbox.

Questions from your reviewer

Write to tech@celorus.com. A person answers, and we will walk your reviewer through how the desk is set up for your firm.