Gemini integration: ask Roadmark from Gemini

Gemini connects to outside systems through the Model Context Protocol (MCP), and Roadmark's connector is an MCP server. What changes is where you add it. Google offers custom MCP servers in two products with different rules: the Gemini app at gemini.google.com, for people on personal Google Accounts, and Gemini Enterprise, for teams.

Which Gemini to use

Gemini Enterprise, for a team. An admin adds Roadmark once, and each person signs in. On the Business edition, a team admin adds it under Manage team > Connected apps, as Google describes in Set up your custom MCP server connection; Google marks that feature as pre-GA. On the Standard, Plus, Pay-as-you-go and Frontline editions, an admin creates a Custom MCP Server data store in the Google Cloud console, following Set up your custom MCP server data store. Both ask for the server's address and its OAuth details, and Gemini Enterprise registers as an OAuth client so that each person signs in on their own. Google says the Gemini Enterprise app doesn't use your prompts and outputs to train its models, in its Gemini Enterprise FAQ.

The Gemini app, for one person. Anyone who meets Google's conditions can add Roadmark as a custom app, following Connect & manage custom apps for Gemini Apps. Google lists them plainly: be 18 or over and in the US, sign in with a personal Google Account (work and school accounts can't add custom apps yet), use it in English, and keep Keep Activity on. That last condition matters for business data: the Gemini Apps Privacy Hub says Google uses that activity to improve its services, including training generative AI models, and that human reviewers read a subset of chats. Read it before connecting company freight data.

What your team can ask Gemini

Anything they could find in Roadmark, asked in their own words. See the AI assistants page for questions by team.

For example, with Roadmark's example group, an owner asks Gemini how many loads Keel Transport hauled for Ashgrove Logistics this quarter, and at what rate. Keel Transport joined the group in March and still dispatches on McLeod, with Roadmark running beside it. Gemini reads what Roadmark receives from McLeod, the loads, status, stops and documents, and answers from those records. The owner can see all three companies because their Roadmark access covers the group; a dispatcher at Keel Transport asking the same question would see only Keel Transport's side.

Two sets of permissions, both enforced

Roadmark's. The connector runs as the person asking and carries exactly their access. A Roadmark admin sets how far assistants may go, from read only, to read and draft, to acting within the rules your team has set for the agents. Work asked for in Gemini goes through the same agents as work asked for in the app, so a dollar limit on a lane, a customer on credit hold or a carrier blocked by compliance applies the same way. See AI controls.

Gemini's. In Gemini Enterprise, Google says a new custom MCP server starts switched off: on the Business edition an admin enables the connection, and on the other editions all of its actions are off until an admin turns on the ones the team may use. By default, Gemini Enterprise also asks the person to confirm each action, unless the tool is marked read only. In the Gemini app, Google requires manual confirmation for any write action, and notes that Gemini can make mistakes. On Roadmark's side, anything a chat prepares waits as a draft for a person, unless your admin has chosen "Act within your rules".

Every action taken through Gemini is on the record in Roadmark, with the person, the assistant and the time.

Using it day to day

In the Gemini app, type @ in a chat and pick Roadmark so Gemini uses it for that prompt. Ask in plain words. When a request needs work done, the result lands in Roadmark as a draft, with the reasons, in the same queue your team approves everything else from. To stop using it in the Gemini app, turn Roadmark off or remove it under Settings > Connected Apps; your Roadmark admin can also turn assistants off for a person, a team or a company at any time.

The same connector works in Claude and ChatGPT too, and the Developers page describes the MCP server behind it for teams building their own tools.