Gemini CLI に MCP server を追加する方法

Gemini CLI is Google's open-source agent for the terminal. MCP servers are set under mcpServers in settings.json, for your user or for one project.

✓ stdio✓ HTTP✓ SSE✓ OAuth対応環境: CLIGemini CLI のドキュメントと照合済み

Gemini CLIにMCPサーバーを追加する

1Open ~/.gemini/settings.json, or .gemini/settings.json in the project.
2Paste the server under mcpServers.
3Remote servers use httpUrl for streamable HTTP.
4Start gemini and run /mcp to see the server and its tools.

例: Codebase Memory

~/.gemini/settings.json

{
  "mcpServers": {
    "codebase-memory-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "codebase-memory-mcp"
      ]
    }
  }
}

Trust and tool filters

trust: true skips confirmation for every tool of that server; keep it for servers you wrote or know well.
includeTools and excludeTools trim what the model sees; when both name a tool, exclude wins.
Sensitive variables from your shell are redacted before they reach a server unless you pass them in env.
OAuth tokens are kept in ~/.gemini/mcp-oauth-tokens.json; /mcp auth signs in again.

Gemini CLI が MCP server を保存する場所

~/.gemini/settings.jsonuser
.gemini/settings.jsonproject

読み込みの確認: /mcp

制限と設定

The default request timeout is 600,000 ms (10 minutes), set per server with timeout.
includeTools and excludeTools filter a server's tools; exclude wins.
trust: true skips confirmations for that server; leave it off unless you trust it.
OAuth needs a local browser; tokens are stored in ~/.gemini/mcp-oauth-tokens.json.

Gemini CLI でサーバーが動かないとき

Run /mcp in a session to see each server's status and tools.
Start with --debug for detailed connection logs.
Use /mcp auth to sign in to an OAuth server again.

Gemini CLIのMCPに関する質問

How do I add an MCP server to Gemini CLI?

Run gemini mcp add <name> <command or URL>, or add an entry under mcpServers in ~/.gemini/settings.json (user) or .gemini/settings.json (project).

Does Gemini CLI support remote MCP servers?

Yes, over SSE and Streamable HTTP, with OAuth sign-in when the server asks for it.

How do I hide some of a server's tools?

List them in excludeTools, or list only the ones you want in includeTools.

Can environment variables be used in the config?

Yes. env values accept $VAR and ${VAR}, and %VAR% on Windows.

Gemini CLI向けの人気MCPサーバー

1
Codebase Memorydeusdata/codebase-memory-mcp4.9万スター
Provides a local knowledge graph that connects source code, documentation, and infrastructure to give AI agents structural context.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "codebase-memory-mcp": { "command": "npx", "args": [ "-y", "codebase-memory-mcp" ] } } }
2
LLM Sandboxvndee/llm-sandbox1143 スター
Securely execute LLM-generated code in isolated containers or local Linux and Windows sandboxes across multiple programming languages.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "llm-sandbox": { "command": "uvx", "args": [ "llm-sandbox" ] } } }
3
Pascal 3D Editorpascalorg/editor2.5万スター
Lets an AI client create, inspect, validate, and save editable 3D building scenes, including walls, rooms, doors, and furnishings.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "editor": { "command": "npx", "args": [ "-y", "@pascal-app/mcp" ] } } }
4
Apify MCP Serverapify/apify-mcp-server1.6万スター
Lets AI agents search, run, and read results from thousands of web scraping and automation Actors on Apify Store.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "apify-mcp-server": { "httpUrl": "https://mcp.apify.com/" } } }
5
CocoIndex Codecocoindex-io/cocoindex-code2981 スター
Provides AST-based semantic code search for local repositories using incremental re-indexing and natural language queries.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "cocoindex-code": { "command": "uvx", "args": [ "cocoindex-code" ] } } }
6
AdLoopklosk/adloop301 スター
Lets an AI client read and change Google Ads, Reddit Ads and GA4 campaigns, with a preview shown before every change.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "adloop": { "command": "uvx", "args": [ "adloop" ] } } }
7
Claude Flowruvnet/claude-flow8.2万スター
An agent meta-harness that provides Claude Code and Codex with specialized agent swarms, self-learning memory, and enterprise security guardrails.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "claude-flow": { "command": "npx", "args": [ "-y", "claude-flow" ] } } }
8
Playwrightmicrosoft/playwright-mcp4万スター
Enables LLMs to interact with web pages through structured accessibility snapshots using Playwright browser automation.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "playwright-mcp": { "command": "npx", "args": [ "-y", "@playwright/mcp" ] } } }
9
Atlassian Dc MCP Bitbucketb1ff/atlassian-dc-mcp109 スター
Allows an AI client to interact with Bitbucket Data Center repositories and code through the Model Context Protocol.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "atlassian-dc-mcp": { "command": "npx", "args": [ "-y", "@atlassian-dc-mcp/bitbucket" ] } } }
10
Next AI Draw.iodayuanjiang/next-ai-drawio3.7万スター
Lets an AI agent draw and edit draw.io diagrams on a live browser canvas that you can also edit by hand.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "next-ai-drawio": { "command": "npx", "args": [ "-y", "@next-ai-drawio/mcp-server" ] } } }
11
Repomixyamadashy/repomix2.9万スター
Packs a local or remote codebase into a single AI-friendly file with token counts so LLMs and coding agents can read or search it.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "repomix": { "command": "npx", "args": [ "-y", "repomix" ] } } }
12
MCP Toolbox for Databasesgoogleapis/mcp-toolbox1.7万スター
Connects AI agents and IDEs to enterprise databases using prebuilt tools or a custom framework for secure, structured queries.
Gemini CLIの設定
~/.gemini/settings.json
{ "mcpServers": { "mcp-toolbox": { "command": "docker", "args": [ "run", "-i", "--rm", "us-central1-docker.pkg.dev/database-toolbox/toolbox/toolbox:1.17.0" ] } } }

Gemini CLIにMCPサーバーを追加する

1Open ~/.gemini/settings.json, or .gemini/settings.json in the project.

2Paste the server under mcpServers.

3Remote servers use httpUrl for streamable HTTP.

4Start gemini and run /mcp to see the server and its tools.

例: Codebase Memory

~/.gemini/settings.json

{
  "mcpServers": {
    "codebase-memory-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "codebase-memory-mcp"
      ]
    }
  }
}

Trust and tool filters

trust: true skips confirmation for every tool of that server; keep it for servers you wrote or know well.

includeTools and excludeTools trim what the model sees; when both name a tool, exclude wins.

Sensitive variables from your shell are redacted before they reach a server unless you pass them in env.

OAuth tokens are kept in ~/.gemini/mcp-oauth-tokens.json; /mcp auth signs in again.

制限と設定

The default request timeout is 600,000 ms (10 minutes), set per server with timeout.

includeTools and excludeTools filter a server's tools; exclude wins.

trust: true skips confirmations for that server; leave it off unless you trust it.

OAuth needs a local browser; tokens are stored in ~/.gemini/mcp-oauth-tokens.json.

Gemini CLIのMCPに関する質問

How do I add an MCP server to Gemini CLI?

Run gemini mcp add <name> <command or URL>, or add an entry under mcpServers in ~/.gemini/settings.json (user) or .gemini/settings.json (project).

Does Gemini CLI support remote MCP servers?

Yes, over SSE and Streamable HTTP, with OAuth sign-in when the server asks for it.

How do I hide some of a server's tools?

List them in excludeTools, or list only the ones you want in includeTools.

Can environment variables be used in the config?

Yes. env values accept $VAR and ${VAR}, and %VAR% on Windows.

Gemini CLI向けの人気MCPサーバー

Codebase Memorydeusdata/codebase-memory-mcp4.9万スター

Provides a local knowledge graph that connects source code, documentation, and infrastructure to give AI agents structural context.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "codebase-memory-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "codebase-memory-mcp"
      ]
    }
  }
}

LLM Sandboxvndee/llm-sandbox1143 スター

Securely execute LLM-generated code in isolated containers or local Linux and Windows sandboxes across multiple programming languages.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "llm-sandbox": {
      "command": "uvx",
      "args": [
        "llm-sandbox"
      ]
    }
  }
}

Pascal 3D Editorpascalorg/editor2.5万スター

Lets an AI client create, inspect, validate, and save editable 3D building scenes, including walls, rooms, doors, and furnishings.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "editor": {
      "command": "npx",
      "args": [
        "-y",
        "@pascal-app/mcp"
      ]
    }
  }
}

Apify MCP Serverapify/apify-mcp-server1.6万スター

Lets AI agents search, run, and read results from thousands of web scraping and automation Actors on Apify Store.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "apify-mcp-server": {
      "httpUrl": "https://mcp.apify.com/"
    }
  }
}

CocoIndex Codecocoindex-io/cocoindex-code2981 スター

Provides AST-based semantic code search for local repositories using incremental re-indexing and natural language queries.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "cocoindex-code": {
      "command": "uvx",
      "args": [
        "cocoindex-code"
      ]
    }
  }
}

AdLoopklosk/adloop301 スター

Lets an AI client read and change Google Ads, Reddit Ads and GA4 campaigns, with a preview shown before every change.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "adloop": {
      "command": "uvx",
      "args": [
        "adloop"
      ]
    }
  }
}

Claude Flowruvnet/claude-flow8.2万スター

An agent meta-harness that provides Claude Code and Codex with specialized agent swarms, self-learning memory, and enterprise security guardrails.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "claude-flow": {
      "command": "npx",
      "args": [
        "-y",
        "claude-flow"
      ]
    }
  }
}

Playwrightmicrosoft/playwright-mcp4万スター

Enables LLMs to interact with web pages through structured accessibility snapshots using Playwright browser automation.

Gemini CLIの設定

~/.gemini/settings.json

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

Atlassian Dc MCP Bitbucketb1ff/atlassian-dc-mcp109 スター

Allows an AI client to interact with Bitbucket Data Center repositories and code through the Model Context Protocol.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "atlassian-dc-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@atlassian-dc-mcp/bitbucket"
      ]
    }
  }
}

Next AI Draw.iodayuanjiang/next-ai-drawio3.7万スター

Lets an AI agent draw and edit draw.io diagrams on a live browser canvas that you can also edit by hand.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "next-ai-drawio": {
      "command": "npx",
      "args": [
        "-y",
        "@next-ai-drawio/mcp-server"
      ]
    }
  }
}

Repomixyamadashy/repomix2.9万スター

Packs a local or remote codebase into a single AI-friendly file with token counts so LLMs and coding agents can read or search it.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "repomix": {
      "command": "npx",
      "args": [
        "-y",
        "repomix"
      ]
    }
  }
}

MCP Toolbox for Databasesgoogleapis/mcp-toolbox1.7万スター

Connects AI agents and IDEs to enterprise databases using prebuilt tools or a custom framework for secure, structured queries.

Gemini CLIの設定

~/.gemini/settings.json

{
  "mcpServers": {
    "mcp-toolbox": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "us-central1-docker.pkg.dev/database-toolbox/toolbox/toolbox:1.17.0"
      ]
    }
  }
}