Legal

Privacy policy

Last updated: October 16, 2026

Overview

mlti.ai (“we”, “us” or “our”) gives you a Chief of Staff and a team of agents. Each agent looks after one area of your work: it reads the mail and calendars you connect, keeps notes, prepares briefs and asks you what it needs. mlti also documents processes: you talk a piece of work through and mlti writes it up. This policy describes what we collect when you use the web app, the browser extension and texting, how we use it, and who helps us run the service.

mlti never uses your mail, your calendar or anything else you put in mlti to train AI models.

By using mlti.ai, you agree to the practices described in this policy. If you do not agree, please do not use the service.

Information we collect

Account information

Your name and email address, and whether you sign in with a password, with Google, with Microsoft or with a code we email you. We also keep which workspaces you belong to and your role in each.

Mail and calendars you connect

When you connect a Google account, agents read the mail and calendar events described under Mail and calendars below, and keep what they read.

What your agents and Chief of Staff write

Agents’ notes, briefs, questions for you and the decisions they record, and your conversations with the Chief of Staff, on the web or by text.

Files and notes you add

Documents, notes and links you give mlti or an agent to work from.

Captures

When you talk a piece of work through with mlti’s voice agent, we keep the transcript of the conversation. If you take it by phone, we also keep the number you gave. When you use the browser extension during a capture, we note the actions you take in your browser: page and tab navigation, the buttons and links you click (including their on-screen labels), the text you enter into forms (with password fields masked), and content you copy or paste. The extension never records video of your screen or raw keystrokes, and it notes actions only while a capture is running. If you turn on screenshots for a capture, it also saves pictures of the page, as described under Screenshots.

Processes

The flow charts, runbooks and other documentation mlti produces, and the comments and sign-offs on them.

Messages to JB

If you use Talk to JB in the app (a chat with mlti’s founder), your message and any picture you attach go to mlti’s Slack, with your name, email, workspace, the page you were on and a link to that session’s replay, so he can answer.

Usage and analytics

We use PostHog, a product-analytics provider, to understand how the service is used: page views, clicks on buttons and links, and where you came from. When you’re signed in, PostHog also receives your email, workspace, role and plan. PostHog also records session replays, a reconstruction of how you moved through our pages that helps us fix problems. Analytics data is processed in the United States.

In the app, a replay keeps the layout of each page and the movement of your pointer, but every word on screen is masked, pictures are blocked, and what you type is never recorded. Analytics events record which button or link was clicked but not its words. Page addresses lose their search terms and share links but keep their path, which can include an agent’s name. Network requests and console logs are not recorded.

On our marketing site (mlti.ai), which shows only our own copy, replays mask what you type.

Mail and calendars

What’s read

Each agent reads mail from one connected Google account, and only the mail that matches what is set for it:

  • the people or domains it was sent from or to
  • words in it
  • how far back to start (on the free plan, at most 30 days)

You set these, as do your workspace’s managers and admins. The agent skips Gmail’s Promotions, Social and Forums tabs. To see what’s coming up today and tomorrow, mlti also reads the main calendar of each Google account you connect: each event’s title, time, attendees and description.

What’s kept, and where

For each email an agent reads, mlti keeps its sender, recipients, date, subject and text, and the attachments it can read (up to 10 MB each) with the text in them. For a meeting mlti matches to an agent, it keeps the meeting’s title, time, attendees and description with that agent. It’s stored in mlti’s database and file storage (Supabase) and made searchable for your agents (Voyage AI). What agents write from it, such as notes, briefs and questions for you, cites the email each fact came from.

Who can see it

Everyone in a workspace can open that workspace’s agents, including their notes, briefs and the emails they read. In a workspace that is only yours, that is only you.

Never used to train AI

mlti never uses your mail or calendar to train AI models. To do an agent’s work, mlti sends the text it works from to the AI providers listed under Service providers.

Disconnecting

You can disconnect a Google account at any time in Settings, under Connected accounts. That stops all reading from it. What agents already read and wrote stays in mlti until it is deleted. You can’t delete an email an agent read from inside mlti: email privacy@mlti.ai and we will delete what you ask for, from one email to an agent with everything it read and wrote. Archiving an agent stops it; mlti keeps what it read and wrote, and Undo brings it back right after.

Google user data

mlti’s use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. In particular, we:

  • use Gmail and Google Calendar data only to provide the features you see in mlti
  • don't sell it, use it for advertising, or use it to train AI models
  • transfer it only to the service providers that help provide those features, or when the law requires it

Texting the Chief of Staff

If you link a phone in Settings to text your Chief of Staff, we keep your phone number and the texts you send and receive over iMessage. They travel through Linq and Apple, so they aren’t end-to-end encrypted between you and mlti. Your texts and the Chief of Staff’s replies are kept in a conversation called Text messages, which you can open in the web chat and delete there. You can unlink the phone in Settings at any time, and we email you when you link or unlink one. Texts from mlti only ever link to mlti.ai.

Your daily brief by text is off until you turn it on, in Settings or by texting your Chief of Staff. Then each morning it texts you a link to your brief, with a headline above it on a day something can’t wait, and never in your quiet hours (20:00 to 08:00 in your timezone, unless you change them). The link works for a day. Its button opens that one brief, and the reply drafts of the questions it lists, read-only, on that device for 24 hours; anyone you forward the link to can do the same. When you ask for a reply draft by text, the text back links to it the same way. In our database we keep only a scrambled copy of the link, of the key your browser holds, and of the address that pressed the button. Unlinking the phone ends its links. You can turn it off at any time.

The Chief of Staff answers iMessage only. A text sent by SMS or RCS gets at most one short reply a day and nothing acts on it. When someone who isn’t a customer yet texts mlti over iMessage, they join the waiting list: they get at most three short replies, and one more when we let them in, and we keep their number, the name they give and what they tell us is on their mind, until we let them in, or for 90 days if we don’t. Anthropic, our AI provider, reads each text, with mlti’s last reply to them, only to choose a reaction. We keep nothing of that beyond what the waiting list keeps. We keep a working copy of each incoming text while we handle it. Once it’s handled, that copy keeps only the number and what happened, not the words, and it’s deleted a week later. The words of your texts and the replies stay only in your Text messages conversation.

Screenshots

Screenshots are optional and off by default. You choose whether to turn them on each time you start a capture, and your browser remembers your last choice.

What’s captured

With screenshots on, the browser extension saves a picture of the browser tab you’re looking at each time you click, and whenever you press its camera button. It saves still pictures only, never video. It doesn’t take one when you click a password or card field, or on the browser’s own pages, such as settings or the Chrome Web Store. A screenshot shows whatever is visible in that tab at that moment, and nothing in it is blurred, so leave screenshots off for a capture that would show information you don’t want in your documentation.

Where it’s kept

In private storage in mlti, with the rest of your workspace’s data. Screenshots don’t have a public address: each one is shown only after we check the viewer’s sign-in or share link.

Who can see it

People in your workspace who can open the process, and anyone you share the process with, by a share link to the process or to its folder. Screenshots appear under the steps of the process’s document.

Deleting them

Deleting a process deletes its screenshots. To delete your account and all of its data, see Data retention below.

Cookies and similar technologies

We use cookies and your browser’s local storage to keep you signed in, to let you open a process someone shared with you or a daily brief from its text, to remember your preferences (such as your theme and the size of panels), and for the product analytics and session replay described above. The analytics cookie is shared across our marketing site (mlti.ai) and the app (app.mlti.ai), so we can understand how people discover and start using mlti. We don’t show a cookie banner.

You can block or delete cookies through your browser settings, though signing in needs them. To object to analytics and session replay, email privacy@mlti.ai.

How we use your information

  • To provide and operate the mlti.ai service
  • To run your agents and Chief of Staff: read the mail and calendars you connect, keep notes, prepare briefs, answer you and ask what they need
  • To write up your captures as processes
  • To let AI apps you connect read your workspace's approved processes, and leave notes and requests when you ask them to
  • To understand how the service is used and fix problems, from usage data
  • To communicate with you about your account or the service
  • To comply with legal obligations

Data retention

We keep your information for as long as your account is active or as needed to provide the service, with these exceptions:

  • The short summary an agent writes about each decision it makes is deleted after 30 days.
  • The working copy of each incoming text loses its words once it's handled, and is deleted a week later.
  • A linked phone's number is kept until you unlink it in Settings, or until you're no longer in the workspace it texts.
  • A daily brief is deleted after 30 days.
  • The link in a daily brief's text is deleted a day after it stops working, and the record of each press of it two days after the press.

You can delete notes and files you added, processes, comments and your conversations with the Chief of Staff in the app. To delete an email an agent read, your workspace or your account, email privacy@mlti.ai and we will delete it, with all associated data.

Sharing and disclosure

We do not sell your personal data. We share it only with the service providers listed below, as far as each needs it to provide its part of the service, or when the law requires it.

AI connections

You can connect an AI app you use, such as Claude, ChatGPT or Cursor, to mlti. The app sends you to mlti to sign in and approve the connection. From then on it reads from mlti as you, and anything it leaves in mlti is in your name.

What a connected AI app can read

  • Your workspace's approved processes: the latest version a manager signed off, with its steps, guardrails, owner, folder, and who approved it and when.
  • The agents your workspace has published: each one's name, what it's for and its address. At an agent's own address, an AI app reads only that agent's charter and the documents linked to it.
  • Your name and email address, which the app receives when you sign in so it knows who is connecting.

It can’t read drafts or changes still in review, capture conversations, files you uploaded, or the mail your agents keep. It can’t read agents’ notes and briefs either, unless a manager signs one off like a process. It can’t change any process.

Notes and requests

At your request, a connected AI app can leave a note on a process, for example when a step no longer matches how the work is done, or request a process that doesn’t exist yet. We store the text of the note or request in mlti with your name, the AI app’s name and the time. A note appears in the process’s comments, where people in your workspace who can open that process can read it, and we email the process’s owner (or, if it has none, whoever signed it off). A request is visible to your workspace’s managers and admins, and we email them. A note or request never changes a process. An admin can turn notes and requests off for the whole workspace.

What we log, and why

We log every request a connected AI app makes: who made it, which AI app, what it did (a search, a read, a note or a request for a process), which process it asked for and which approved version it read, and when. We log which process was read, not what was asked: the words of a search aren’t kept, only that a search happened. We use this log to show your workspace how often its AI reads each process, to answer which version an AI read on a given day, to limit how many requests an app can make, and to investigate problems and misuse. We keep it for as long as your workspace’s account is active. We also record that a request happened, without its content, in the analytics and monitoring tools listed below.

Disconnecting

You can disconnect an AI app at any time, in mlti under Settings, then Your AI, or in the AI app’s own connector settings. An admin of your workspace can turn AI connections off for the whole workspace.

What the AI provider does with it

What a connected AI app reads goes to the company that provides it, such as Anthropic, OpenAI or Cursor. We don’t control how that company stores or uses it. Its own terms and privacy policy apply.

Service providers (subprocessors)

We rely on these third-party service providers to operate mlti.ai. Each receives only what it needs to provide its part of the service, under its terms with us (Exa, Deepgram and ElevenLabs under their terms with Composio or Vapi). We do not sell your data to them.

  • Supabase: Database, sign-in and file storage: your account, everything agents and the Chief of Staff keep and write, your processes, uploaded files, email attachments and screenshots
  • Vercel: Hosting for the app and this website, background jobs, and reading the text out of Word, Excel and iPhone photo files you upload or that arrive as email attachments
  • Anthropic: The AI models behind agents, the Chief of Staff, captures and the reactions mlti puts on texts. They receive the text they work from, including mail, calendar text and texts sent to mlti
  • Voyage AI: Makes what agents keep searchable. It receives the text agents keep, including mail, and the words they search for
  • Composio: Connects the accounts you choose (Google and Notion), reads Gmail and Google Calendar for your agents, sends a process to Notion when you confirm it, and runs agents' web searches
  • Exa: Web search, through Composio, when an agent looks something up. It receives the search words
  • Inngest: Background processing: runs agents' tasks and writes up your captures
  • Vapi: The voice agent in a capture, and the phone call if you take it by phone
  • Deepgram: Turns what you say in a capture into text, through Vapi
  • ElevenLabs: The voice agent's spoken replies, through Vapi
  • TypeSafe AI: Gives each decision an agent records its altitude. It receives a short summary the agent wrote about that decision, which can quote a line from an email, and summaries of up to six other decisions made for the same person, each with the altitude that person gave it
  • Linq: Carries the iMessages between your phone and your Chief of Staff, if you link a phone
  • Slack: Where messages you send through Talk to JB arrive (see Messages to JB), and where we're told when someone signs up, joins a workspace or the waiting list, starts a trial, subscribes or buys voice time, with the person's name and email or the workspace's name
  • Stripe: Payments
  • Resend: Email from mlti, such as invitations and account notices
  • PostHog: Product analytics and session replay (described above)
  • Axiom: Application logs and performance monitoring

Security

We host your data with Supabase and Vercel, which have SOC 2 Type II reports, and they encrypt it in transit and at rest. Inside a workspace, what each person can open and change depends on their role.

Your rights

Depending on your location, you may have rights to access, correct, delete, or export your personal data. To exercise any of these rights, contact us at privacy@mlti.ai.

If you are in the European Economic Area or the United Kingdom, you have rights under the GDPR and UK GDPR, including the right to object to processing based on our legitimate interests. We rely on our legitimate interest in understanding and improving the service for analytics and session replay, and you may object at any time by contacting privacy@mlti.ai.

Changes to this policy

We may update this policy from time to time. We will notify you of material changes by email or via a notice in the application. Your continued use of mlti.ai after changes take effect constitutes your acceptance of the revised policy.

Contact

For privacy-related questions or requests: privacy@mlti.ai