✨ Gemini MCP Setup
Before You Start
- Power User subscription — Get one here
- JT Power Tools extension installed in Chrome — Install
- Gemini CLI installed — Install from GitHub
- A JT Power Tools account — you sign in with Google or an email link, the same way you sign in to the portal
- An AI key for your JobTread org (an owner or admin adds one on the portal's AI connections page)
Setup Instructions
Open Gemini Settings
Edit your Gemini CLI settings file at:
On Windows, this is typically %USERPROFILE%\.gemini\settings.json. Create the file if it doesn't exist.
Add the MCP Server Config
Add the following to your settings.json file under the mcpServers key. To tie the connection to one JobTread org, use that org's MCP URL from its card on AI connections instead:
Sign In
Restart Gemini CLI, then run /mcp auth jobtread. Your browser opens.
Sign in on the JT Power Tools page. Your AI app opens it for you. Sign in the way you sign in to the portal: Google or an email link. If your company has AI keys for more than one JobTread org, pick one. An org's own MCP URL (from AI connections) skips this. Then click Allow.
Start Using It
That's it! Ask Gemini about your JobTread data. Try:
"Show me my active jobs" or "What tasks are due this week?"
Gemini-Specific Notes
Gemini CLI uses the httpUrl key (not url) for remote MCP servers. This is per Google's Gemini CLI documentation. The JT Power Tools extension generates the correct format automatically when you select Gemini.
Needs a browser: Gemini CLI signs you in through a browser on the same computer. Over SSH or in a container with no browser, use the headless setup at the bottom of this page.
Gemini CLI vs gemini.google.com: MCP is currently supported through the Gemini CLI (command line tool). The gemini.google.com web interface does not yet support connecting custom MCP servers directly.
How It Works
Gemini CLI connects to the JT Power Tools MCP server over OAuth. You sign in once, on the JT Power Tools page, the way you sign in to the portal. Gemini CLI keeps an access token after that — it never sees your sign-in or your company's AI key.
The MCP server provides a full set of tools for working with your JobTread data — jobs, budgets, tasks, schedules, contacts, daily logs, and more. Tools cover reading, writing, and deleting. What each connection can do is governed by the permission tier (Read, Write, or Delete) an admin sets for the grant key.
Troubleshooting
"Can't connect"? The page says why and what to do: join your team, have an owner or admin add an AI key on AI connections, or have an owner upgrade on Plan & billing. If it says the link expired, go back to your AI app and connect again.
Link expired, or Gemini CLI says it isn't signed in — Run /mcp auth jobtread again to sign in fresh.
"MCP server not found" — Make sure you're using httpUrl (not url) in your Gemini settings. Restart the CLI after saving.
"Grant Key expired" — Grant keys expire after 3 months of inactivity. Generate a fresh one in JobTread and have an owner or admin swap it in with Replace key on its card on AI connections.
"Unavailable while role mirroring is on" — Your connection uses a License Key and Grant Key, which role mirroring denies. Remove the server and add it again with just the URL — you'll sign in, and your JobTread role applies.
Still stuck? Join our Discord for help.
Connecting without sign-in (headless / CI)
Use this only where nobody can open a browser, like a CI job or a server. Everywhere else, the sign-in above is simpler and keeps keys out of your config.
Add your License Key and Grant Key as a header. Your License Key is in your subscription email; the Grant Key is your JobTread API grant key. Each org's card on AI connections shows its MCP URL (use it if its address differs from the URL below), with the key masked, so use the full grant key you created in JobTread.
Save settings.json and restart Gemini CLI. A 401 means a key is wrong: check both keys and the Bearer LICENSE_KEY:GRANT_KEY format, with a colon between them.