What we collect, and what never leaves your machine.

What never leaves your computer

The daemon does not upload your files, your source code, your environment variables, your API keys or your Claude, Codex or Cursor credentials. It reads them locally, because the agent needs them, and it sends none of them to us.

What it does send is the transcript of a run: the messages in the room, the tool calls the agent made and their results as the agent reported them, the request status, and basic machine metadata (a machine name, the operating system, which runtimes are installed, whether the machine is online).

A tool result can quote a file the agent read. That quoted text is part of the timeline and is visible to the room, which is the point of the timeline. If you do not want a room to see something, do not ask the agent to open it in that room.

What we collect

Account data: your email address, your display name and your avatar, from Google sign-in or from an emailed sign-in code. We store the email code attempt count to rate-limit abuse.

Room data: messages you send, attachments you upload, agent replies and output files, the tool-call timeline, approvals, membership and invitations.

Machine data: the machines you pair, their names, operating systems, installed runtimes, online status, the IP address a machine connects from and the country that address resolves to. We show you the IP so you can recognise your own machine and notice one you do not.

Integration data, if you connect one: for Slack, the workspace and channel identifiers we need to deliver a message back to the right conversation.

Operational logs: standard request logs and error reports from our hosting provider.

What we do not do

We do not sell your data. We do not share it with advertisers. We do not train models on your rooms, your transcripts or your files, and we do not host models at all — your prompts go to the vendor whose subscription your agent already uses.

We do not run third-party advertising or analytics trackers on muti.im.

Who processes it for us

Vercel hosts the web application. Supabase hosts the Postgres database. Cloudflare R2 stores room attachments, agent output files and blog images. An SMTP provider delivers sign-in codes and notification email. Google provides optional sign-in. If you connect Slack, Slack receives the messages you route through it.

These are processors: they handle data on our instructions to run the service, not for their own purposes.

Cookies

We set a session cookie so you stay signed in, and a language cookie that remembers which of the twelve site languages you chose. That is all. No advertising cookies, no cross-site tracking, and no consent banner to dismiss, because there is nothing to consent to.

How long we keep it

Rooms, messages and timelines are kept while your account is active, so that a request from six weeks ago still has an audit trail. Attachments and output files live in object storage on the same basis.

When you delete a room, a machine or a membership, we mark it deleted and remove it from every view immediately. Deleted records are purged from backups on our ordinary backup rotation.

Ask us to delete your account and we remove your account data and the rooms you own.

Your rights

You can ask for a copy of your data, ask us to correct it, or ask us to delete it. Email service@muti.im from the address on the account and we will act within thirty days.

muti is in private beta and the service changes often. When a change affects what we collect, we update this page and change the date at the top rather than announce it only in a changelog.

Children

muti is a tool for work and is not directed at children under 16. We do not knowingly collect their data.

Contact

Muti AI — service@muti.im. Privacy questions, deletion requests and security reports all reach a person at that address.