Ask Gemini

Community

Bridge Claude with Gemini CLI for AI-to-AI review and second opinions

How to install ↓ Compare GitHub source repository ↗

Category: Dev Tools #192 of 438
Stars: ★ 18 Very popular
Transport: stdio (local process)
Runtime: Node.js 20+
Credentials: API key / credential required
License: MIT
Last commit: today
Tools: 5

55FMRS · C

Ask Gemini gives users who already have enterprise Gemini CLI access a convenient way to obtain a second model's opinion inside clients like Claude, with 1M+ token context and @ file syntax as its main strengths. However, since 2026-06-18 the Gemini CLI is enterprise-only, so free and individual subscription accounts cannot use it, sharply limiting its audience. Code also leaves the machine, and the project is unofficial and third-party, so data handling and account eligibility should be evaluated first.

Strongest · Setup experience 12/20 Weakest · Reliability 9/20

Reliability

9/20

Security and permissions

11/20

Maintenance

11/20

Documentation

12/20

Setup experience

12/20

Why each score

Reliability 9/20: Only the manifest and README were available; the actual source of packages/gemini-mcp/server.json, its tool schemas, any CI workflow, and any tests were not provided. Consequently the declared tools (ask-gemini, ask-gemini-edit, fetch-chunk, get-usage-stats, ping) cannot be confirmed to match real behavior, and the stdio init plus tools/list handshake cannot be verified. The README does describe configurable failure surfaces such as GMCPT_TIMEOUT_MS (default 300000 ms) and GMCPT_LOG_LEVEL, but that is self-reported evidence only. Under the static calibration rule, without verifiable execution evidence (real CI plus committed tests covering key paths) reliability is capped at 12; missing evidence is a deduction, so 9.
Security and permissions 11/20: The server launches an external Gemini CLI over stdio and accepts @ file syntax to ingest the codebase, so user source code is transmitted to Google's model service; the README notes Gemini CLI is enterprise-seat-only since 2026-06-18. The project does offer Ollama as a local, private alternative, and both manifest env vars (GMCPT_TIMEOUT_MS, GMCPT_LOG_LEVEL) are marked non-secret and optional, with no hardcoded tokens or real secrets in install examples. However, the third-party data flow lacks demonstrated per-call confirmation and concrete data-flow disclosure, and least privilege cannot be verified from the available evidence. Scored as the middle band, main risks visible but permission/confirmation/scoping incomplete: 11. No red-line indicators (malware, credential theft, covert exfiltration) were found, so the 0-4 band does not apply.
Maintenance 11/20: MIT-licensed and not archived, 18 stars with 11 open issues; manifest version 1.2.1 matches npm package @ask-llm/gemini-mcp 1.2.1, and the README documents a migration from the legacy ask-gemini-mcp name to the @ask-llm scope, indicating some release and versioning discipline. Static evidence shows no CHANGELOG, dependency-update policy, security-response channel, commit history, or issue-response latency, and seven MCP servers sharing one repository dilutes maintenance attention. Governance and update-path gaps remain, so active-but-incomplete: 11.
Documentation 12/20: Documentation is broad and multi-provider: install steps for Claude Code, Cursor, Codex CLI, Claude Desktop, Pi and generic STDIO; a tool-to-package table; disclosed limits such as Gemini 1M+ context, the enterprise restriction, fallback conditions (quota for Gemini/Codex, overload for Claude, rate limit for Antigravity), and default timeout/log env vars; plus an explicit non-affiliation disclaimer. Gaps: the docs focus on the unified @ask-llm/mcp package, while parameter-level detail, error taxonomy and troubleshooting specific to the scored single-provider packages/gemini-mcp/server.json are thin, and README and manifest are untrusted evidence that cannot be cross-checked against the implementation. Deducted to 12.
Setup experience 12/20: The install path is clear: Node.js 20+ on Linux/macOS, provider prerequisite npm install -g @google/gemini-cli && gemini login, then a one-line registration via npx -y @ask-llm/gemini-mcp with config examples for several clients. Concrete obstacles remain: Gemini CLI is enterprise-only since 2026-06-18, so free, Google AI Pro and Ultra accounts get no output and the README itself recommends Antigravity/Codex/Claude/Ollama instead, making the default path unusable for most users; Windows support is not stated and appears unsupported. Under static calibration setup is capped at 15 absent verifiable execution evidence, and the practical gate justifies 12.

Static review · not runReviewed 2026-09-25

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesRuns commands or codeUses the network

Best for

Large-context, whole-codebase code review
Teams that already hold enterprise Gemini Code Assist seats
Developers who want to consult Gemini inside clients like Claude or Codex
File-scoped prompts using @ syntax for precise references

Not for

Users with only free, Google AI Pro, or Ultra accounts (Gemini CLI is enterprise-gated since 2026-06-18)
Environments where the Gemini CLI cannot be installed or authenticated
Users expecting zero extra CLI setup

Required permissions

Requires a locally installed and authenticated Gemini CLI
Communicates with the local MCP client over stdio
Reads referenced code file contents included in prompts and sends them to Gemini

Risks and side effects

Code content leaves the machine for Gemini's service; assess data-transfer risk for sensitive code
Non-enterprise accounts cannot use Gemini CLI and may only receive guidance instead of results
The default 5-minute timeout may be exceeded on large whole-repo requests; adjust GMCPT_TIMEOUT_MS
Unofficial third-party project, not supported by Google or Anthropic

Setup

Before you start

Runtime：Node.js 20+

XAI_API_KEY optionalsecret Key for the xAI (Grok) API, obtained from the xAI console; needed only when using Grok features.

Other optional settings (2)

GMCPT_TIMEOUT_MS optional Timeout for Gemini CLI execution in milliseconds; default 300000 (5 minutes).

GMCPT_LOG_LEVEL optional Log verbosity: debug, info, warn, or error; default warn.

Install Node.js 20+ (Linux or macOS).
Install and log in to the Gemini CLI: npm install -g @google/gemini-cli && gemini login (requires an enterprise Gemini Code Assist seat).
Register the server in your client: for Claude Code run claude mcp add --scope user gemini -- npx -y @ask-llm/gemini-mcp; for Cursor add it to ~/.cursor/mcp.json or .cursor/mcp.json; for Codex CLI add it to ~/.codex/config.toml; for Claude Desktop add it to claude_desktop_config.json; any other STDIO client uses {"command": "npx", "args": ["-y", "@ask-llm/gemini-mcp"]}.
Optional environment variables: GMCPT_TIMEOUT_MS (default 300000 ms) and GMCPT_LOG_LEVEL (debug/info/warn/error, default warn).

.mcp.json

{
  "mcpServers": {
    "gemini": {
      "command": "npx",
      "args": ["-y", "@ask-llm/gemini-mcp"],
      "env": {
        "GMCPT_TIMEOUT_MS": "300000",
        "GMCPT_LOG_LEVEL": "warn"
      }
    }
  }
}

Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "gemini": {
      "command": "npx",
      "args": [
        "-y",
        "@ask-llm/gemini-mcp"
      ],
      "env": {
        "GMCPT_TIMEOUT_MS": "300000",
        "GMCPT_LOG_LEVEL": "warn"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add gemini -e GMCPT_TIMEOUT_MS=300000 -e GMCPT_LOG_LEVEL=warn -- npx -y @ask-llm/gemini-mcp

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

The client's tool list should show ask-gemini, ask-gemini-edit, fetch-chunk, get-usage-stats, and ping; calling the ping tool and getting a successful connection response confirms the setup.

Troubleshooting

Confirm Node.js version 20+
Run gemini login to check CLI authentication and gemini --version to confirm installation
If you get guidance rather than output, check whether your account has an enterprise Gemini Code Assist seat
Raise GMCPT_TIMEOUT_MS for very large codebases
Set GMCPT_LOG_LEVEL to debug for detailed logs
Call the ping tool to verify the connection

Things to try

Once connected, you can ask your AI assistant things like:

ask gemini to summarize @. the current directory
ask gemini to review src/auth.ts for security issues
ask gemini to debate the plan in docs/design.md
show me the token usage stats for this session

Tools 5

ask-gemini read-only

Ask Gemini via the Gemini CLI with @ file syntax, 1M+ token context, and live progressive output via stream-json

ask-gemini-edit writes

Get structured OLD/NEW code edit blocks from Gemini

fetch-chunk read-only

Retrieve chunks from cached large responses

get-usage-stats read-only

Per-session token totals, fallback counts, and breakdowns by provider/model (in-memory only)

ping read-only

Connection test

Use cases

Have Gemini review a code diff and surface issues the primary model missed

Read and summarize an entire codebase in one call using 1M+ token context

Ask about specific files or directories with @ file syntax

Get a second opinion on an architecture or design plan

Obtain structured OLD/NEW edit suggestions via ask-gemini-edit

Supported clients

Claude Code

Claude Desktop

Cursor

Codex

Listed from the project's documentation, not tested by this site.

Overview

Ask Gemini is the Gemini-only local MCP server from the Lykhoyda/ask-llm repository (package @ask-llm/gemini-mcp, version 1.2.1). It uses stdio transport to connect existing MCP clients to the Gemini CLI, letting Claude Code, Claude Desktop, Cursor, Codex CLI, and other clients ask Gemini to review a diff, evaluate a plan, or give an independent second opinion. Its main draws are Gemini's 1M+ token context and @ file syntax, which allow ingesting a whole codebase in one call, plus live progressive output via stream-json. The repository also ships Codex, Claude, Grok, Antigravity, and Ollama servers, which are not part of this server. The source states clearly that since 2026-06-18 the Gemini CLI is enterprise-only (Gemini Code Assist Standard/Enterprise seats); free, Google AI Pro, and Ultra accounts lost access, and non-enterprise accounts get actionable guidance instead of output. This is an unofficial third-party tool, not affiliated with Anthropic, Google, OpenAI, or xAI, and is MIT licensed.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.4k · Tools 2 Compare with this →

NPM Sentinel MCP Server 79 · B

AI-powered NPM package analysis MCP server

★ 18 · Tools 19 Compare with this →

MartinLoop MCP Server 79 · B

An independent control layer that gives AI coding agents budgets, brakes, and receipts

★ 191 Compare with this →

GitHub MCP Server 76 · B

GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language

★ 33.2k · Tools 15 Compare with this →

Compare with other MCP servers

More servers from this repository

Data synced 2026-09-25 Read the FMRS scoring method