Google · Connection guide

Terminal

Connect Gemini CLI
to Love CRM.

Gemini CLI can connect to Love with an HTTP MCP server entry. Add your workspace URL, start /mcp auth love-crm in Gemini CLI, and complete Love’s teammate consent flow. The client can then discover the CRM tools exposed to that teammate.

HTTP transport and OAuth sign-in are documented in Gemini CLI’s MCP guide.

See setup steps Visit Gemini CLI

Setup · Hosted MCP

From your workspace
to your agent.

Before starting, have an enabled Love workspace, Owner authorization and an active AI teammate with the right permissions for the work. The example URL is a placeholder; copy the actual address from Love’s API & MCP setup.

  1. Register Love’s HTTP endpoint

    Run the command below with your actual workspace MCP address. This is a remote Streamable HTTP connection, rather than a local process.

  2. Open the authentication flow

    Start Gemini CLI and run /mcp auth love-crm. Use the browser sign-in flow offered by the client.

  3. Choose the teammate in Love

    Follow the browser sign-in, authenticate as a Love Owner, and select an active AI teammate. Review its role, capabilities and assigned actions before consenting. The connection uses that teammate’s identity, rather than inheriting your Owner access.

  4. Check the connection before doing work

    Ask the agent to identify the company and bound teammate, list permitted tools and assigned actions, and perform a read-only lookup. Confirm the result before requesting a business action.

Gemini CLI · then run /mcp auth love-crm inside Gemini
gemini mcp add --transport http love-crm https://your-workspace.lovecrm.co/mcp

Client setup reviewed . Official Gemini CLI documentation

A client’s always-allow setting does not grant additional Love capabilities. Keep the Love teammate’s role narrow enough for the work you intend to delegate.

Make the first request
a connection check.

Use this read-only prompt before delegating work. A successful sign-in should identify the right company and teammate, not just show a connected icon.

Read-only connection check
Identify the Love company and AI teammate connected to this session.
Call get_mcp_connection and get_bound_teammate.
Use list_catalog_actions with the assigned scope, then inspect get_tenant_catalog.
Summarize the permissions and action types you actually received.
Do not submit proposals, apply routes, send messages, change records or collect payments.

Useful questions
to ask next.

Build a repeatable morning briefing from permitted customer and appointment records, or investigate service-round questions from the terminal. Keep factual lookups separate from proposals that change the business.

  • Summarize the upcoming appointments this Love teammate can read, and link the records behind the answer.
  • List the teammate’s permitted catalog actions before proposing an office workflow.

Use only the tools actually returned for the connected teammate. For a catalog business action, ask for a proposal and review it in Love’s Home workspace. Routing collaboration follows its own workflow-specific permissions and apply controls.

Gemini CLI + Love

Your connection questions,
answered.

Which Gemini CLI transport should I choose?

Use http for Love’s hosted /mcp endpoint. Gemini CLI distinguishes Streamable HTTP from legacy SSE and local stdio servers.

Why are tools missing after sign-in?

Check the server’s connection status, Gemini’s tool filters and the chosen Love teammate’s permissions. A connection does not guarantee every CRM tool is available to that teammate.

Can I connect an AI client that is not listed here?

A compatible client needs remote Streamable HTTP MCP and Love’s OAuth authorization flow. Add the exact workspace URL, authorize an active AI teammate, and start with read-only discovery. Support for local MCP servers or API keys alone does not establish compatibility. Confirm your client’s current documentation and workspace policy before using it.

What is Love CRM’s MCP address?

Use the exact URL in your workspace’s API & MCP setup. The address follows https://your-workspace.lovecrm.co/mcp; replace your-workspace with the provisioned workspace hostname. A public URL identifies the endpoint but does not authorize access.

Does connecting an agent expose my whole business?

The hosted connection is bound to a selected active AI teammate in one tenant. Its role permissions and assigned actions limit what it can use. Review that role carefully: broad read permissions can still expose broad business context to the external agent provider.

Can the connected agent grant itself more permissions?

The teammate connection cannot use an Owner acting role, change users or teammate governance, or approve its own catalog action proposals. Love applies the permission checks in its API, beyond instructions supplied to the model.

What requires human approval?

Business actions submitted through Love’s catalog proposal path wait for review in Home. Workflow-specific tools, including routing collaboration, have their own capabilities and execution paths. Review those separately; a connection is not a blanket promise that every tool uses the same approval process.

How do I stop a connection?

Revoke the connection’s authorization or deactivate its bound Love teammate. Love rechecks the teammate and connection state on subsequent requests. Removing a client configuration alone is different from revoking its server-side authorization.

Is setup available to every Love account?

Love is in early access. Use these guides with an enabled workspace, an active teammate and a compatible client. Confirm tenant availability and client-specific support with Love before treating a connection as ready for your business.

See Love with your agent

More ways to work

One business.
Your choice of agent.

xAI · Personal agent

Grok Bot

Workspace MCP setup for Grok Bot.

Read connection guide

Meta · Muse & Muse Code

Muse

Muse Code setup and personal Muse compatibility.

Read connection guide

OpenAI · Desktop, CLI & IDE

Codex

Workspace MCP setup for Codex.

Read connection guide

Anthropic · Web & desktop

Claude

Workspace MCP setup for Claude.

Read connection guide

Anthropic · Terminal

Claude Code

Workspace MCP setup for Claude Code.

Read connection guide

Cursor · Editor & agent

Cursor

Workspace MCP setup for Cursor.

Read connection guide

OpenAI · Web · custom plugin

ChatGPT

Workspace MCP setup for ChatGPT.

Read connection guide