Your AI
Connect mlti to your AI.
Your AI finds and follows the processes your managers approved, the newest approved version. Connect it once, and every process you sign off is there the next time someone asks how your team does it.
How it works
Signed off in mlti. Followed by your AI.
A manager signs off a process in mlti. That approved version is what your AI reads.
Someone asks their AI how your team does something, like closing the month or handling a refund.
Their AI searches your approved processes, reads the right one, and tells you whose sign-off it's following.
When a process changes, capture it again. Once the new version is approved, your AI reads it on the next request. Nothing to download or upload.
mlti never runs anything. It holds the steps your team signed off. Your AI does the work, in your own account.
Things to ask once you’re connected
“How do we onboard a new client here?”
“Walk me through our month-end close.”
“A client wants a refund for last week's visit. What's our process?”
When something’s off
Your AI tells the owner. The owner decides.
When a step doesn’t match what’s in front of you, or there’s no process at all, your AI can let the right people in mlti know. Claude asks you before it sends either one.
A step doesn't match
“This step says QuickBooks, but we use Xero now.”
Your AI leaves a note on that step for the process's owner. It shows in the process's comments in mlti as an AI note, with your name and the app it came from, and the owner gets an email.
There's no process for it
“How do we handle a chargeback?”
If nothing approved matches, your AI can request one. Managers and admins see it in mlti under “Requested by your team's AI”, get an email, and can capture it or dismiss it.
Neither one changes a process. The owner reads the note and decides what to do, and your AI keeps reading the approved version until a manager signs off a new one. Admins can turn notes and requests off in Settings, then Your AI.
Set it up
Paste one address. Sign in once.
You need an mlti account and at least one approved process. Each person connects as themselves, so their AI reads only what they can read in mlti.
The address
https://app.mlti.ai/mcpYour AI sends you to mlti to sign in and approve. The approval screen shows where the connection is going, such as claude.ai, so you know it’s the AI you meant.
Claude Pro or Max
You set it up yourself.
- In Claude, open Settings and go to Connectors.
- Choose to add a custom connector. Name it mlti and paste the address above.
- Click Connect, sign in to mlti, and approve.
To try it, ask Claude how your team handles something you've captured in mlti.
Claude Team or Enterprise
An owner adds it once. Then everyone connects.
- A Claude owner opens your organization's settings in Claude and goes to Connectors.
- They add a custom connector named mlti with the address above. This happens once, for the whole team.
- Then each person finds mlti in their own Connectors settings, clicks Connect, signs in to mlti and approves.
Everyone signs in as themselves, so Claude reads only what that person can read in mlti.
Claude Code
Connect on claude.ai, and it follows you.
- Connect mlti on claude.ai first, using the steps for your Claude plan.
- Sign in to Claude Code with the same Claude account. Connectors you add on claude.ai show up there too.
- To check, type /mcp in Claude Code and look for mlti in the list.
Adding mlti from inside Claude Code doesn't work yet, so start on claude.ai.
ChatGPT
Uses developer mode.
- In ChatGPT's settings, turn on developer mode. Whether you can depends on your plan, and on a business plan an admin may need to allow it.
- Create a new connector (ChatGPT may call it an app). Name it mlti, paste the address above, and choose OAuth if it asks how to sign in.
- When ChatGPT opens mlti, sign in and approve.
In a chat, make sure mlti is turned on before you ask about one of your processes.
Cursor
Added as an MCP server.
- Open Cursor's settings and find the MCP section.
- Add a new MCP server named mlti with the address above. If Cursor opens its mcp.json file, add the lines below.
- When Cursor asks you to log in to mlti, sign in and approve.
{
"mcpServers": {
"mlti": {
"url": "https://app.mlti.ai/mcp"
}
}
}What your AI can see
Approved processes. Nothing else.
Your AI can
- Read the processes your managers approved in your organization, the newest approved version.
- See each process's owner and folder, who approved it and when, and its audit id, so it can tell you whose sign-off it's following.
- Leave a note for a process's owner, or request a process that's missing, with your OK.
- See your name and email address, so it knows who's connecting.
Your AI can’t
- See drafts, or changes that are still in review.
- See capture conversations, recordings or files you uploaded.
- Change any process. Notes and requests go to people in mlti, who decide what changes.
The newest approved version. While a newer version of a process is in review, your AI keeps reading the last approved one and says a newer one is on the way. If a manager revokes a sign-off, the process comes off your AI right away.
Everything your AI does in mlti is logged. mlti records who connected, which AI app, what it did (a search, a read, a note or a request), which process, and when, plus the version for every read. It logs what was read, not what was asked: the words of a search aren’t kept. Each process in mlti shows how often your team’s AI has read it.
Your AI provider’s terms apply to what it reads. Once your AI reads a process, the company behind it (Anthropic, OpenAI or Cursor) handles it under its own terms. Our privacy policy has the details.
Disconnect
Turn it off any time.
In mlti
Open Settings, then Your AI. You’ll see each AI app you’ve connected and can disconnect any of them. Admins see everyone’s connections and can turn AI connections off for the whole organization, or turn off just notes and requests.
In your AI app
Remove mlti from the app’s connector settings. In Claude, that’s Settings, then Connectors. Your AI stops reading from mlti straight away.
Help
If something isn’t working
Your AI only sees approved processes. Open the process in mlti and check that it's approved. If it is, ask again and name the process, or ask your AI to look it up in mlti. If there's no process for it yet, your AI can request one.
Remove mlti from your AI's connector settings, add it again with the address https://app.mlti.ai/mcp, and approve on the mlti screen. Sign in with the mlti account that belongs to your organization.
Someone disconnected it in mlti, or you signed out of mlti, which disconnects your AI apps too. Sign back in to mlti, then connect again from your AI.
An admin in your organization switched AI connections off in mlti, under Settings, then Your AI. Ask them to turn it back on.
An admin in your organization switched off notes and requests in mlti, under Settings, then Your AI. Your AI can still read approved processes. Ask an admin to turn them back on, or tell the process's owner yourself.
Connect on claude.ai instead. Connectors you add there show up in Claude Code when you sign in with the same Claude account.
Each AI app can make 60 requests a minute for each person. Wait a minute and ask again.
Still stuck? Email info@mlti.ai and tell us which AI you’re using. Privacy questions go to privacy@mlti.ai.
Not using mlti yet? Start with one process.
Talk it through once and get it signed off. From then on, your AI follows it.
7 days free. No card.