Extension or settings.json: which should I use?

The extension is one command and adds the LeadMind server, a context file that teaches Gemini how LeadMind works, and four /leadmind: slash commands. Edit settings.json yourself if you only want the tools or need a per-project setup.

Why httpUrl and not url?

In Gemini CLI settings, httpUrl means a Streamable HTTP server, which is what LeadMind runs. The url field is for the older SSE transport and will not connect to LeadMind.

Should I set trust to true?

Leave it off. With trust off, Gemini CLI asks before each tool call, so you see create_lead_job before a job holds credits. Turning trust on skips every confirmation for that server.

Does this work with the Gemini API as well?

Yes, differently: the Gemini API uses function calling rather than MCP. LeadMind publishes a manifest of function declarations for it; the MCP setup guide shows a minimal executor.

LeadMind in Gemini CLI

Gemini CLI can call LeadMind’s MCP server from your terminal: build a scored lead list, verify an email file or scan a local market without leaving the shell. Install the LeadMind extension for slash commands, or add the server to settings.json yourself.

Setup guide · Updated 28 September 2026

Start free Watch the demo

Option 1: the LeadMind extension

gemini extensions install https://github.com/tafveez-omorfiagroup/leadmind-gemini-extension

Restart Gemini CLI and run /mcp auth leadmind to sign in. The extension adds these commands:

/leadmind:lead-list: Turns a brief into one precise LeadMind prompt, runs the job, answers its clarifying question, and returns a top-10 table with a CSV link.
/leadmind:verify-emails: Chooses a quick check (up to 25) or a bulk job (up to 5,000), checks your quota first, and explains risky and unknown results honestly.
/leadmind:shortlist: Ranks delivered leads by score and reachable channel, then drafts one opener per lead grounded in what LeadMind found. Drafts only.
/leadmind:market-scan: Samples a category in an area and reports the share of businesses with WhatsApp, online booking, a given platform or an outdated site.

Option 2: add the server yourself

From the shell:

gemini mcp add --transport http leadmind https://outreach-nlfc.onrender.com/mcp

Or add it to ~/.gemini/settings.json (every project) or .gemini/settings.json (this project only):

{
  "mcpServers": {
    "leadmind": {
      "httpUrl": "https://outreach-nlfc.onrender.com/mcp",
      "timeout": 60000
    }
  }
}

Start Gemini CLI and run /mcp auth leadmind. Sign in to LeadMind in the browser and approve access.
Run /mcp to confirm the server is connected and its tools are listed. gemini mcp list shows the same from the shell.
Ask for leads, or use a slash command if you installed the extension.

Using an API key instead of OAuth

On a server or in CI, add a header with a key from Settings:

{
  "mcpServers": {
    "leadmind": {
      "httpUrl": "https://outreach-nlfc.onrender.com/mcp",
      "headers": { "Authorization": "Bearer lp_live_…" }
    }
  }
}

Your first request

Try this first

“/leadmind:market-scan dental clinics in Jumeirah”

Gemini samples the category and reports the share of clinics with WhatsApp, online booking or an outdated site. Without the extension, ask the same question in plain words.

The tools Gemini gets

Tool	What it does
`create_lead_job`	Start a live lead job from a sentence; credits are held, charged only on delivery.
`get_job_status`	Progress, ETA, chosen source and credits held while the job runs.
`clarify_job`	Answer the job’s clarifying question; same job, no new credits.
`get_leads`	Read delivered leads with contacts, score, reason and website signals.
`export_leads_csv`	A 24-hour download link for the full CSV.
`get_credit_balance`	Balance, credits held, and the key’s daily cap.
`list_sources`	The public sources LeadMind routes between.
`verify_emails`	Check up to 25 addresses inline: deliverable, risky, undeliverable or unknown.
`start_email_verification`	Verify a list of up to 5,000 as a background job.
`get_email_verification`	Progress and counts of a bulk job, then a CSV of every result.
`get_verification_quota`	Checks used and left this billing period.

Troubleshooting

The server shows as disconnected. Check the entry uses httpUrl, not url, then run /mcp auth leadmind again.
Calls time out while a job runs. Tool calls return quickly and the job keeps running on LeadMind’s side; ask Gemini to check the job status. A timeout of 60000 ms leaves room for slow networks.
Slash commands are missing. They come with the extension; check gemini extensions list.

Using the Gemini API from your own code instead? The MCP setup guide covers the function-calling manifest.

Frequently asked questions

Extension or settings.json: which should I use?: The extension is one command and adds the LeadMind server, a context file that teaches Gemini how LeadMind works, and four /leadmind: slash commands. Edit settings.json yourself if you only want the tools or need a per-project setup.
Why httpUrl and not url?: In Gemini CLI settings, httpUrl means a Streamable HTTP server, which is what LeadMind runs. The url field is for the older SSE transport and will not connect to LeadMind.
Should I set trust to true?: Leave it off. With trust off, Gemini CLI asks before each tool call, so you see create_lead_job before a job holds credits. Turning trust on skips every confirmation for that server.
Does this work with the Gemini API as well?: Yes, differently: the Gemini API uses function calling rather than MCP. LeadMind publishes a manifest of function declarations for it; the MCP setup guide shows a minimal executor.