An MCP (Model Context Protocol) server + standalone agent SDK for the Google Ads API.
Built by googleadsagent.ai · MIT License
# From PyPI (once published)
pip install googleadsagent-mcp
# From GitHub right now
pip install git+https://github.com/itallstartedwithaidea/google-ads-mcp.git
# With uv
uv add googleadsagent-mcp
# From source
git clone https://github.com/itallstartedwithaidea/google-ads-mcp.git && cd google-ads-mcp && pip install -e .| Client | Transport | Config |
|---|---|---|
| Claude Code | stdio | .mcp.json in project root (included) |
| Claude Desktop | stdio | claude_desktop_config.example.json |
| Cursor / Windsurf | stdio | Settings → MCP → python -m ads_mcp.server |
| OpenAI Agents SDK | stdio | MCPServerStdio(command="python", args=["-m", "ads_mcp.server"]) |
| LangChain | stdio | langchain-mcp-adapters |
| Remote / Cloud | HTTP SSE | python -m ads_mcp.server --http |
# 1. Copy and fill credentials
cp .env.example .env
# 2. Validate environment
python scripts/validate.py
# 3a. Run as MCP server (stdio — Claude Desktop/Cursor/Claude Code)
python -m ads_mcp.server
# 3b. Run as MCP server (HTTP — remote agents)
python -m ads_mcp.server --http
# 3c. Run as standalone agent CLI
python scripts/cli.py
python scripts/cli.py --single "Show campaign performance for account 1234567890"Place .mcp.json in your project root (already included):
{
"mcpServers": {
"google-ads": {
"type": "stdio",
"command": "python",
"args": ["-m", "ads_mcp.server"],
"env": {
"GOOGLE_ADS_CREDENTIALS": "${GOOGLE_ADS_CREDENTIALS}",
"GOOGLE_ADS_LOGIN_CUSTOMER_ID": "${GOOGLE_ADS_LOGIN_CUSTOMER_ID}"
}
}
}
}Claude Code auto-discovers .mcp.json. CLAUDE.md is written specifically for Claude Code
to orient itself — it reads this first on every session.
Add to your project's .gemini/settings.json:
{
"mcpServers": {
"google-ads": {
"command": "python",
"args": ["-m", "ads_mcp.server"],
"env": {
"GOOGLE_ADS_CREDENTIALS": "/path/to/google-ads.yaml",
"GOOGLE_ADS_LOGIN_CUSTOMER_ID": "1234567890"
}
}
}
}For the full Gemini CLI setup with Google Ads slash commands and agent skills, see gemini-cli-googleadsagent.
Remote option — if you deploy the buddy-agent on Cloudflare Workers, you can skip local credentials entirely and connect via SSE:
{
"mcpServers": {
"buddy-google-ads": {
"url": "https://your-buddy-agent.workers.dev/sse"
}
}
}from agents.mcp import MCPServerStdio
import os
server = MCPServerStdio(
command="python",
args=["-m", "ads_mcp.server"],
env={
"GOOGLE_ADS_CREDENTIALS": os.environ["GOOGLE_ADS_CREDENTIALS"],
"GOOGLE_ADS_LOGIN_CUSTOMER_ID": os.environ.get("GOOGLE_ADS_LOGIN_CUSTOMER_ID", ""),
}
)
# All 29 Google Ads tools automatically availablelist_accessible_customers · list_accounts · execute_gaql · get_campaign_performance
get_keyword_performance · get_search_terms · get_ad_performance · get_account_budget_summary
generate_keyword_ideas
get_auction_insights · get_change_history · get_device_performance · get_geo_performance
get_recommendations · get_pmax_performance · get_impression_share
update_campaign_budget · update_campaign_status · update_ad_group_status · update_keyword_bid
add_keywords · add_negative_keywords · remove_negative_keyword · create_campaign
create_ad_group · switch_bidding_strategy · generic_mutate
get_gaql_reference · get_workflow_guide
# Option 1: google-ads.yaml path (recommended)
GOOGLE_ADS_CREDENTIALS=/path/to/google-ads.yaml
GOOGLE_ADS_LOGIN_CUSTOMER_ID=123-456-7890
# Option 2: individual env vars
GOOGLE_ADS_DEVELOPER_TOKEN=...
GOOGLE_ADS_CLIENT_ID=...
GOOGLE_ADS_CLIENT_SECRET=...
GOOGLE_ADS_REFRESH_TOKEN=...
GOOGLE_ADS_LOGIN_CUSTOMER_ID=123-456-7890Full credential setup guide: google-ads-api-agent README → Step 1A
Google Ads API v23 has 70+ services. Current status:
- ✅ 29 tools implemented — core reporting, audit, write operations, doc references
- 🔧 42 services planned — full roadmap in
docs/SERVICES.md - ⬜ ~20 out of scope — platform admin, deprecated, LSA-specific
google-ads Python client library — Copyright 2023 Google LLC
https://github.com/googleads/google-ads-python
All API calls in this package use thegoogle-adspip package published by Google. The v23 service definitions, proto types, and gRPC clients are Google's work.
googleads/google-ads-mcp — https://github.com/googleads/google-ads-mcp
Patterns: MCP singleton,search_stream+field_mask.paths
google-marketing-solutions/google_ads_mcp — https://github.com/google-marketing-solutions/google_ads_mcp
Patterns: doc-serving MCP tools, remote OAuth,omit_unselected_resource_names
cohnen/mcp-google-ads — https://github.com/cohnen/mcp-google-ads
Patterns:format_valueproto serialization helper
gomarble-ai/google-ads-mcp-server — https://github.com/gomarble-ai/google-ads-mcp-server
Patterns: dual transport (stdio + HTTP), KeywordPlanIdeaService
itallstartedwithaidea/google-ads-api-agent — https://github.com/itallstartedwithaidea/google-ads-api-agent
Patterns: auction insights, change history, campaign creator, bidding strategy manager, negative keywords, geo targeting, PMax reporting, Filter-First Architecture
Full per-feature attribution table: docs/SERVICES.md
- google-ads-skills — Anthropic Agent Skills for Claude (analysis, audit, write, math, MCP)
- google-ads-api-agent — Full Python agent with 28 API actions and 6 sub-agents
- google-ads-gemini-extension — Gemini CLI extension with 22 MCP tools, skills, commands, and themes
- googleadsagent.ai — Production deployment (Buddy) on Cloudflare with semantic memory, billing, and monitoring
MIT — see LICENSE
google-ads dependency: Apache 2.0, Copyright 2023 Google LLC
FastMCP: Apache 2.0