Lia is not locked inside Leaf360.
The stack
Foundation models
Claude, GPT, Gemini. General intelligence.
Leaf360
The AI Operating System for mortgage. Your pipeline, your lenders, your rules.
You
Loan officers, teams and wholesale partners. The one driving.
The MCP server is the line between the top two. Open protocol, your permission, your workspace.
What the Leaf360 MCP server is
The Model Context Protocol is an open standard for connecting an AI assistant to a system it does not own. An MCP server exposes what a product can do, and any MCP client can then use it, with the user's permission.
The Leaf360 MCP server is our side of that. Point Claude, ChatGPT, Cursor or anything else that speaks MCP at your Leaf360 workspace and you can work on your own leads, loans, partners, tasks, templates and automations from inside that assistant, without copying anything between windows.
This is the shape of the whole company in one feature. Foundation models are general intelligence. Leaf360 is the mortgage layer that turns it into pipeline work. They are complements, and an open protocol is how you prove that rather than just say it.
What people actually do with it
Not a demo trick. These are the jobs it gets used for, in order.
Clean up and import your data
Point your assistant at the export from your old CRM. It reads the mess, works out what maps to what, and gets the file into shape before it ever reaches the importer. This is the job people ask for most.
Set Leaf360 up for yourself or your team
Stages, pipelines, custom fields, lead sources, team structure. Describe how your business actually runs and have it built, instead of clicking through settings screens one at a time.
Build automations
Design and create workflows from the assistant you already have open. Workflows themselves are a Core and Pro feature, so this one follows your plan.
Design email templates anywhere
Draft a template in Claude Design, GPT or Claude and drop it straight into Leaf360. Ask for the merge variable list first, or the copy will not merge correctly.
Ask about your own pipeline
Your leads, your loans, your partners and your tasks, answered inside the assistant you were already working in, without another tab.
Bring it into your own tooling
Any MCP client works, present or future. If your team already lives in one, Leaf360 shows up there rather than asking everyone to move.
Which clients work
MCP is an open standard, so this list grows without us shipping anything.
Claude
Desktop, web and Claude Code.
ChatGPT
Through its connector support for MCP.
Cursor
For anyone building alongside their pipeline.
Gemini CLI
Google's command-line agent documents MCP server support. Gemini CLI specifically, not every Gemini app.
Any MCP client
The protocol is open. If it speaks MCP and supports the transport and authentication, it can connect.
Connect in a few minutes
One address, no key. Pick your assistant, paste the address where it asks for a server, and sign in with your own Leaf360 login when the browser opens. The assistant then works your workspace with the same permissions you have.
1.Choose your assistant
2.Copy the connection address
https://mcp.leaf360.com/mcpThe same address for everyone. Nothing to generate and no key to paste: when your assistant asks you to authorize, you sign in with your own Leaf360 login, and the assistant gets the permissions your login has and no more.
3.Connect and authorize in Claude
Web and desktop app. Free, Pro, Max, Team and Enterprise; Free allows one custom connector.
Connected to Leaf360 this way
- 1
Open Settings, then Connectors.
- 2
Click the plus, then Add custom connector, and paste the address above.
- 3
Click Add, then Connect, and sign in with your Leaf360 login when the browser asks. On Team and Enterprise an owner adds it under Organization settings first, then each person clicks Connect.
Pro, Plus, Business, Enterprise and Education, on the web. Custom servers are a developer-mode feature.
Steps from the client's official guide, not yet tried with Leaf360
- 1
Open Settings, then Security and login, and turn on Developer mode. On a workspace plan an admin turns it on first.
- 2
Open the apps settings, press the plus, and paste the address above to create the app.
- 3
Sign in with your Leaf360 login when ChatGPT sends you to authorize. The app appears under Drafts until you use it.
Terminal. A connector added in Claude on the web also shows up here when you sign in with the same account.
Steps from the client's official guide, not yet tried with Leaf360
- 1
Run the command below once.
- 2
Type /mcp inside Claude Code and follow the browser sign-in, or run claude mcp login leaf360.
claude mcp add --transport http leaf360 https://mcp.leaf360.com/mcpFor anyone building alongside their pipeline.
Steps from the client's official guide, not yet tried with Leaf360
- 1
Put the block below in ~/.cursor/mcp.json for every project, or in .cursor/mcp.json for one.
- 2
Open Cursor's MCP settings, find leaf360 and approve the sign-in when it prompts you.
{
"mcpServers": {
"leaf360": { "url": "https://mcp.leaf360.com/mcp" }
}
}Google's command-line agent. This is the Gemini client with documented MCP support, not the consumer Gemini app.
Steps from the client's official guide, not yet tried with Leaf360
- 1
Put the block below in your Gemini CLI settings.json.
- 2
Run /mcp auth leaf360 inside Gemini CLI and sign in with your Leaf360 login. The CLI discovers the sign-in flow on its own.
{
"mcpServers": {
"leaf360": { "httpUrl": "https://mcp.leaf360.com/mcp" }
}
}OpenAI's terminal agent.
Steps from the client's official guide, not yet tried with Leaf360
- 1
Run the two commands below. The second opens the browser sign-in.
codex mcp add leaf360 --url https://mcp.leaf360.com/mcp
codex mcp login leaf360MCP is an open standard, so this list is not the whole list.
Steps from the client's official guide, not yet tried with Leaf360
- 1
Add a remote server with the address above. The client needs to speak streamable HTTP and handle a sign-in flow it discovers from the server; most current clients do.
- 2
When it asks you to authorize, sign in with your Leaf360 login. There is no key to paste.
- 3
If your client cannot complete the sign-in, tell support which client it is and we will look at it with you.
4.Check the connection
Tested in the Leaf360 demo workspace
Start with a read-only check.
After connecting, ask for one loan with the total count, or one lead with the total count. Confirm that the result belongs to the Leaf360 workspace you intended to authorize before requesting any change.
In the fictional demo workspace on 15 September 2026, a loan search limited to one record with the total count returned one loan and a total of 169, and the same lead search returned one lead and a total of 171. With explicit authorization, a separate test then created one clearly labeled demo task on one demo lead, listed it as pending, completed it, and listed it as completed. Those counts describe that demo at that moment; they are not marketing figures, and this was one connector, one workspace and one task, not a test of every tool, client or an approval card.
First prompt
Using the connected Leaf360 tools, confirm which workspace I am connected to and list the tasks you can help me with. Start with read-only checks. Do not send messages or create, edit or delete records. If you are not connected, tell me which setup step is missing instead of guessing.
What success looks like
- The assistant names your workspace and the kinds of records it can see: leads, loans, contacts, partners, tasks, templates and automations.
- It lists what it can do without changing anything, and it did not send or create anything to find out.
If it does not
- It names a different workspace. You signed in with a login that belongs to another workspace. Disconnect, then connect again with the login for the workspace you want.
- It says it cannot see leads or loans. Your Leaf360 role does not include them. Ask your company admin; the assistant can only see what your own login can see.
- It says it is not connected. The sign-in did not finish. Run the client's connect or auth step again and complete the browser sign-in.
- It shows no Leaf360 data at all. Confirm the intended Leaf360 account was authorized, then follow that client's reconnect steps. Removing the connector on the client side is that client's step; it is not described here as revoking anything on the Leaf360 side.
5.Try a useful task
Each of these reads or drafts first and waits for you before anything is sent or created. They are starting points; none of them was part of the demo verification above.
Pipeline summary
Give me a read-only summary of my pipeline: how many leads are in each stage, which loans are closest to closing, and which leads have gone quiet the longest. Do not change anything.
Touches: Reads leads and loans only.
Draft a follow-up
Help me draft a follow-up for a lead I choose. Ask me to identify the lead, use only information available in my connected workspace, and show me the draft for review before anything is sent.
Touches: Drafts only. Nothing goes out until you send it yourself.
Plan a CRM import
I am moving from another CRM. Ask me for the column headers in my export, tell me which Leaf360 lead and loan fields exist, and propose a mapping with anything that will not map. Do not create or import any records.
Touches: Reads field lists only.
Plan an automation
I want a workflow that creates a task for me when a lead goes quiet for 7 days. Check which triggers and actions are available, show me the steps you would build, and wait for my go-ahead before creating anything.
Touches: Reads capabilities. Building the workflow follows your plan.
The help center carries the same steps, and support will walk you through it if a client will not complete the sign-in.
Reading about Leaf360 from an assistant instead of connecting one? The plain-text summary of this site is at leaf360.com/llms.txt, with the full version at llms-full.txt.
Why we opened it
Every other system in this category is closed. Their AI works on their screens and nowhere else, because their business depends on you staying inside the walls.
We do not have an LOS to protect. Leaf360 connects to the systems you already run, and Lia works in the assistant you already have open. If we are the best place to do the work, you will stay because of that.
- The same agent works in Leaf360 and in Claude, on the same data.
- Your import cleanup happens in the tool that is good at cleanup, then lands in Leaf360.
- A template designed in Claude Design drops straight in, merge variables and all.
- It is on every plan for a limited time, including Free.
Mortgage AI starter kit
Practice organizing contacts and drafting follow-up with fictional data, in ChatGPT, Claude or Gemini, before you connect anything.
Meet Lia
What Lia does inside Leaf360, and what agentic means in practice.
Integrations
Every system Leaf360 connects to, by category.
All features
Everything Leaf360 does, grouped, with the plan each one starts on.
Questions about the MCP server
It is a Model Context Protocol server that connects an AI assistant to your Leaf360 workspace. MCP is an open standard for letting an assistant work with a system it does not own. With it connected, Claude, ChatGPT, Cursor or any other MCP client can work on your leads, loans, partners, templates and automations directly, with your permission, instead of you copying things between windows.
Use Lia where you already work.
Start free, connect the assistant you already have open, and give her your old CRM export to clean up.
Free plan. No credit card. Published pricing.