Skip to main content
Your Agent Builder doesn’t have to live in one browser tab. Connect the AI you already talk to all day, and it gets the keys to the same builder: it can see your agents, draft new ones, launch them in Test mode, take them live, and tidy them up. Everything it builds shows up in your dashboard with a little tag, like via Claude, via Grok Bot or via Cursor (MCP), so you always know who made what.

The one thing you need

It’s also on Settings → Developers → MCP in your dashboard, with a Copy button.

​
Pick your AI

1

Add a custom connector

In Claude, open Settings → Connectors → Add custom connector. Name it Lawtte and paste the server address. Leave advanced settings empty.
2

Sign in

Claude shows a connect card the first time it needs your account. Sign in to Lawtte and pick your firm.
3

Say hi

What agents does my firm have?
More detail: Connect Claude.
Firms whose agents Lawtte runs for them connect once Lawtte switches the connector on. If the MCP page says it isn’t on yet, ask your Lawtte account team.

​
Your first five minutes

Try these in order. Each one is a real thing your AI can do now.
What agents do we have, and what tools and inboxes are connected?
It reads your agents, the systems you’ve connected (Clio, GHL, Slack…), your inboxes and text lines, and your house rules.
Build an agent that follows up with the manufacturer when we ask for the signed release and they go quiet for 5 business days. Up to 3 reminders, CC Karina.
It drafts the blueprint and the wording. You’ll see it in Agents right away, tagged with your AI’s name.
Launch it in Test mode and show me what it would pick up from the last two weeks.
Test mode sends nothing. You get a practice run on real emails so you can see exactly what it would say, and to whom.
Looks great. Take it live.
Your AI asks you to confirm first, then it’s on and sending by itself.
Change the wait to 7 days and pause it over the holidays.
It shows you a before and after for each change, and applies them when you say so.
From now on, never text clients after 7pm, and always CC Karina on manufacturer emails.
It saves these as house rules, and every agent it builds after that follows them. You’ll see them in Agents → Settings → House rules too.
Get us a 916 number that shows OCONNOR LAW on caller ID.
It sends the request to the Lawtte team, who usually set it up within a business day. Then your call agents can use it.
Use the adjuster call agent to call Jane Doe at (916) 555-0142. She agreed to be called and doesn’t have her own lawyer.
It checks the call is allowed (consent, calling hours, Do Not Call), asks you to confirm, then dials and reports back what it learned.

​
The safety rails stay on

Your AI drives, but Lawtte keeps the guardrails:
  • It asks before anything that sends. Going live, running a live agent, applying changes and removing an agent all need your OK in the chat.
  • Test mode first, always. Every new agent starts there.
  • Calls follow the calling rules. Consent, calling hours, Do Not Call and daily limits are checked by Lawtte on every call, and the chat can’t turn them off. See limits and safety.
  • It only sees your firm. One sign-in covers one firm.

​
See what it built, or unplug it

Open Settings → Developers → MCP. You’ll find:
  • Connected apps, each with a Disconnect button. Access ends right away.
  • Built over MCP: everything your AIs have built, with who asked and the status.
Your agents keep running after you disconnect. Pause them in Agents if you want them stopped too.

Writing your own integration, or are you an AI agent?

The exact tool order, config formats and rules.