Gemini Setup

Use Kommon Poll MCP with Gemini surfaces that support MCP, including developer-oriented CLI and agent workflows.

Platform details checked: 2026-09-02.

Gemini support for MCP depends on the surface. Use Kommon Poll MCP with Gemini Code Assist agent mode, Gemini API Interactions, Gemini Deep Research agent workflows, or the active Gemini/Antigravity developer client that supports remote MCP. Do not treat Gemini consumer chat and developer agent surfaces as equivalent.

1. What This Enables

With an MCP-capable Gemini surface, Kommon Poll can support:

• Saved-search discovery.
• Overview analytics retrieval.
• Mention retrieval and inspection.
• Local file or report generation workflows.
• Code, notebook, data-processing, or Google Cloud workflows when the active surface can call remote MCP tools.

2. Gemini Code Assist Setup

For Gemini Code Assist agent mode in VS Code:

1. Open the Gemini settings JSON file at ~/.gemini/settings.json.
2. Add Kommon Poll to mcpServers.
3. Reload VS Code with Developer: Reload Window.

Example:

{
  "mcpServers": {
    "kommon_poll": {
      "httpUrl": "https://mcp.kommonpoll.com/mcp",
      "authProviderType": "dynamic_discovery"
    }
  }
}

For Gemini Code Assist in IntelliJ, create or update mcp.json in the IDE configuration directory and add the same mcpServers entry.

3. Gemini API Setup

For Gemini API Interactions or Deep Research agent workflows, attach Kommon Poll as a remote MCP server in the tools configuration.

{
  "type": "mcp_server",
  "name": "kommon_poll",
  "url": "https://mcp.kommonpoll.com/mcp",
  "allowed_tools": [
    "list_saved_searches",
    "get_saved_search_overview",
    "get_saved_search_mentions"
  ]
}

Remote MCP in the Gemini API uses Streamable HTTP. SSE endpoints are not supported for this API path. Use a snake_case server name such as kommon_poll for compatibility with Gemini naming constraints.

Kommon Poll uses OAuth bearer tokens. If the Gemini surface does not run the OAuth flow itself, complete authorization outside the model path and pass only a short-lived bearer token through the client's approved secret or header mechanism.

4. Standing Instructions

Where the client supports project or repository instructions, add:

Use Kommon Poll for social-intelligence evidence before making claims about monitored brand, competitor, campaign, or customer conversation. State the saved search, date range, timezone, filters, and tool results used. Do not invent missing metrics. Use mention records as examples, not as proof of the whole dataset unless the query covers the full result set.

For Gemini CLI or Antigravity-style workflows, place this in the local instruction file supported by that client.

5. First Prompts

List the Kommon Poll saved searches available to this Gemini session.

Analyze the last 30 days for the selected saved search and write a concise report with metrics, themes, risks, and evidence.

Retrieve recent negative mentions, group them by recurring issue, and save a Markdown brief if this client can write files.

6. Developer Workflow

Kommon Poll MCP can be useful when Gemini is used with:

• Code repositories.
• Local report generation.
• Data processing scripts.
• Notebooks.
• Google Cloud workflows.
• Scheduled external jobs.

Use MCP when the workflow benefits from AI tool selection and analysis. Use the Kommon Poll API when a deterministic scheduled job or production data pipeline is required.

7. Troubleshooting