Gemini CLI MCP server

How it works

Gemini CLI is Google's open source AI agent for the terminal. It speaks MCP, so after a one-line settings change it can reach your blogr account: Search Console performance, the topical map, scheduled and published articles. It signs in with OAuth on its own the first time it needs to.

Gemini CLI acts as you, with your team role on each website, and can change only what your role allows.

What you need

• A blogr account. Each website's data is available while its owner has an active plan or trial.
• Gemini CLI, installed and signed in with your Google account.

Connect Gemini CLI

1. Open ~/.gemini/settings.json (or .gemini/settings.json in a project) and add blogr under mcpServers:

{
  "mcpServers": {
    "blogr": {
      "httpUrl": "https://blogr.ai/mcp"
    }
  }
}

2. Start Gemini CLI and sign in. It opens blogr in your browser, where you choose Allow access:

/mcp auth blogr

3. Run /mcp to check that blogr is connected, then ask "Using blogr, how did my blog do this month?"

Use httpUrl, not url. In Gemini CLI, url means the older SSE transport.

Things to ask Gemini

• "Summarise my Search Console numbers for the last 28 days and list the five queries that grew most."
• "Which silo in my topical map has the fewest live pages?"
• "Show the titles waiting for approval and decline any that repeat an article I already have."
• "Export my published article URLs to a CSV in this folder."

What Gemini CLI can do

Gemini CLI reads without asking. Every tool that changes something is marked as a write, so Gemini CLI asks you to confirm before it runs.

What Gemini CLI cannot do

• Publish, delete or archive articles. Publishing stays in blogr, so nothing reaches your live site from Gemini CLI.
• Start writing an article early. Approved titles are written on their scheduled day, within your monthly allowance.
• See or change billing, website settings, team members, or your publishing connection and its tokens.

An edit to a published article is saved in blogr. The live page changes the next time you publish that article from blogr.

Team roles

Gemini CLI follows your role on each website. Viewers can read everything they see in the app, and Gemini CLI cannot change anything for them. Editors and admins can approve, decline, add, reschedule and edit through Gemini CLI. Owners can do all of it.

Limits

• 120 requests per minute per account, across every assistant you connect.
• 200 changes per account per hour.
• 25 titles added through an assistant per website per day.

Past a limit, Gemini CLI gets a message saying so and can try again later. Nothing you do in the app counts toward these limits.

Your data

• Gemini CLI receives only what you ask for: your websites, articles, topical map, Search Console performance and opportunities.
• blogr never sees your Gemini CLI conversations. For each request it records which tool ran, on which website, whether it worked and how long it took, and deletes that record after 90 days.
• Gemini CLI signs in with its own access token, which never includes your password. To disconnect, remove blogr from mcpServers in your settings.json.

Once Gemini CLI receives data from blogr, Google handles it under its privacy policy. Read blogr's own privacy policy for what we store.

Troubleshooting

• /mcp shows blogr as disconnected: run /mcp auth blogr, finish sign-in in the browser, then restart Gemini CLI.
• Sign-in works on a local callback, so run it on the machine where Gemini CLI runs, not over SSH.
• "blogr needs an active plan or trial": start a plan or trial on the account that owns the website.
• "Your role does not allow this": ask the website owner to give you the Editor or Admin role.

Questions