MCP Setup

Connect Claude, ChatGPT, Cursor, or Gemini CLI to Locari — and manage listings, applicants, and communication directly from your AI assistant.

MCP (Model Context Protocol) is an open standard that lets AI assistants use external services — here Locari — like a toolbox. Once connected, your assistant can retrieve applicant lists, edit listings, schedule viewings, and request documents without you opening the Locari interface.

What is MCP? A protocol that allows AI assistants to communicate with external services. Instead of copy-pasting between a browser and a chat window, the assistant sends requests directly to Locari and receives structured data in return.

What Locari provides over MCP: Read and write access to listings, applicants, documents, viewings, notes, and communication in your account. What you decide: Which assistant you grant access to. You can disconnect any connection at any time under Settings → Integrations.

Common Tasks

Connect Claude
Connect ChatGPT
Connect Cursor, Windsurf, GitHub Copilot, or Cline
Connect Gemini CLI
Disconnect a connection
Review active connections

How To

Connect Claude

Open Settings → Integrations → AI Integrations.
Expand the Claude section.
In Claude Desktop open: Settings → Connectors → Add custom connector.
Name: Locari, URL: paste the displayed MCP URL — your browser opens for sign-in.

What Claude can do after connecting: e.g. "Show me all applicants for Musterstr. 12 with a score above 80" — Claude fetches the current evaluation list directly from Locari. The Schmidt family, score 91/100, appears right in the chat. What you do manually: Confirm accept or reject — decisions are not executed automatically by Claude.

Connect ChatGPT

Open Settings → Integrations → AI Integrations.
Expand the ChatGPT section.
In ChatGPT open: Settings → Connections → Add new MCP server.
Paste the MCP URL — your browser opens for sign-in.

Connect developer tools

For Cursor, Windsurf, GitHub Copilot, and Cline you use a JSON configuration file.

Open Settings → Integrations → AI Integrations.
Expand the desired client and copy the displayed JSON block.
Paste the configuration into the respective settings file:
- Cursor: Settings → MCP → Add Server → Remote
- Windsurf: Cascade → Model Context Protocol → Configure
- GitHub Copilot: create .vscode/mcp.json in your workspace
- Cline: Sidebar → MCP Servers → Advanced Settings

Sample configuration (Cursor / Windsurf / Cline):

{
  "mcpServers": {
    "locari": {
      "type": "url",
      "url": "https://locari.ai/api/mcp"
    }
  }
}

Connect Gemini CLI

Open Settings → Integrations → AI Integrations.
Expand Gemini CLI and copy the configuration.
Paste it into ~/.gemini/settings.json. Gemini CLI uses httpUrl instead of url.

Disconnect a connection

Open Settings → Integrations → AI Integrations.
Locate the connection in the table.
Click the disconnect icon — access is revoked immediately.

Disconnecting a connection cannot be undone. The assistant loses all access rights immediately. You can set up a new connection again at any time.

Views and Fields

Connection table (AI Integrations)

Client name — name of the assistant that authorized the connection
Last used — relative timestamp (e.g. "2 hours ago")
Connected on — creation date of the connection
Disconnect — revokes OAuth access immediately

What the assistant can see and do

The following action areas are available over MCP:

Search and overview — query and filter listings, applicants, and properties
Applicant management — retrieve evaluations, request documents, answer clarifications
Communication — send messages to applicants and team members
Viewings — create and reschedule appointments, send invitations
Documents — manage folders, extract content
Listings — create and edit rental cases, trigger portal export
Notes and tasks — create internal notes and reminders

What Locari checks automatically before an assistant writes: For certain actions (e.g. sending an acceptance), Locari requests an explicit confirmation before the action is carried out. What you decide: Every destructive or final action requires your approval in the chat.

Audit and History

Every action an AI assistant executes over MCP is recorded in the history of the affected object (applicant, listing, document) with the actor "Locari AI" and a timestamp — identical to actions taken from the web interface.

Retention: History follows the same GDPR retention periods as all other activity. Disconnecting an MCP connection does not delete historical entries.

Permissions and Multi-Select

Set up a connection: any team member for their own user account.
Disconnect a connection: only the owner of that connection — takes effect immediately and is irreversible.
Execute actions via the assistant: rights correspond to the role of the signed-in user — a member without admin rights cannot perform admin actions through an assistant either.
View all active connections: Integrations overview — disconnect connections, check status.