mcpField
G
gsearch

gsearch

Local
github.com/daanielcruzSearch4Updated 5d ago

About gsearch

gsearch is a Go-based MCP server that returns Google Search-grounded answers with inline citations. It connects AI tools to Google Search grounding through OAuth or a Gemini API key.

Key features

  1. 01Google Search-grounded answers
  2. 02Inline source citations
  3. 03OAuth or API-key authentication
  4. 04Automatic retry with backoff

Use cases

  1. 01Get current web information in an AI tool
  2. 02Receive cited search results
  3. 03Retry searches after rate limits

MCP support

The server exposes a `google_search` tool over MCP using stdio. MCP-compatible clients can connect it through the configured command.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"daanielcruz-gsearch": {
"command": "npx",
"args": [
"-y",
"@daanielcruz/gsearch-mcp"
],
"env": {
"GSEARCH_PROJECT": "<GSEARCH_PROJECT>",
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"GSEARCH_API_KEY": "<GSEARCH_API_KEY>"
}
}
}
}
Run commandnpx -y @daanielcruz/gsearch-mcp
Environment variables to set
GSEARCH_PROJECTOverride auto-detected Google Cloud project ID (OAuth mode)
GEMINI_API_KEYSecretGemini API key (enables direct API key mode, bypasses OAuth)
GSEARCH_API_KEYSecretAlternate Gemini API key env var (same purpose as GEMINI_API_KEY)

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npm@daanielcruz/gsearch-mcp1.2.3

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README
🚨 Disclaimer 🚨

This is just a heads-up. We haven't seen any bans from GSearch use.

Not affiliated with Google. Uses the same public OAuth client and API as the Antigravity CLI.


GSearch - Free Google Search MCP

GSearch - Free Google Search MCP

Go License: MIT MCP

Give Claude Code, Codex CLI, Cursor, Otherside, and other AI tools real-time web search powered by Google Search grounding. Single binary, zero runtime dependencies.

Install

curl -fsSL https://raw.githubusercontent.com/daanielcruz/gsearch-mcp/main/install.sh | bash

Or build from source:

git clone https://github.com/daanielcruz/gsearch-mcp && cd gsearch-mcp
make build && ./gsearch-installer

The installer downloads the binary and configures Claude Code, Codex, Cursor and Otherside automatically (if available).

How It Works

Claude Code / Codex CLI / Cursor / Otherside / Others
        | MCP (stdio)
   gsearch-server (Go binary)
        | OAuth2
   Google Code Assist API
        | googleSearch grounding
   Answer with [1][2][3] source links

When an AI tool calls google_search, GSearch returns a grounded answer with inline citations:

The current time in Sao Paulo is 08:13 AM.[1][2]
Sao Paulo observes Brasilia Time (BRT), UTC-3.[3]

Sources:
[1] [Time in Sao Paulo](https://...)
[2] [World Clock](https://...)
[3] [Time Zone Info](https://...)

Pricing & Limits

GSearch uses the same Google Search grounding API as Antigravity CLI through its public OAuth client, or the Gemini API with an API key. Limits, availability, and billing depend on the selected authentication method and Google account.

The official Gemini CLI states that free and Google One accounts migrated to Antigravity CLI on June 18, 2026, so GSearch now signs in through Antigravity's public OAuth client. Existing Gemini CLI credentials remain readable.

Rate limits are generous for normal usage. The server retries automatically with dynamic backoff on rapid bursts.

Authentication

Two options:

  1. Google sign-in (Antigravity): the installer opens Antigravity CLI's public OAuth flow with PKCE S256 and a localhost loopback callback. Credentials are stored in ~/.gsearch/oauth_creds.json.
  2. API key: paste a Gemini API key from aistudio.google.com into the installer, or set GEMINI_API_KEY / GSEARCH_API_KEY. This skips OAuth and does not require a project. Google sign-in always takes precedence: when OAuth credentials exist, these variables are ignored.

Antigravity sign-in resolves the Code Assist project automatically (SMS verification may be required on first use). If the account requires a user-defined project, the installer accepts an existing Project ID and retries. It does not create projects or enable billing. The default backend is daily-cloudcode-pa.googleapis.com; set GSEARCH_ANTIGRAVITY_HOST=prod to use cloudcode-pa.googleapis.com.

Configuration

The installer configures everything automatically. For manual setup, add to your MCP client config:

{
  "mcpServers": {
    "gsearch": {
      "command": "npx",
      "args": ["-y", "@daanielcruz/gsearch-mcp"]
    }
  }
}

Works with Claude Code (~/.claude.json), Cursor (~/.cursor/mcp.json), Otherside (~/.otherside/.mcp.json), and other MCP-compatible tools.

For Codex CLI (~/.codex/config.toml):

[mcp_servers.gsearch]
command = 'npx'
args = ['-y', '@daanielcruz/gsearch-mcp']

Set GSEARCH_PROJECT env var only if auto-detection fails.

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.