Connecting the Google Keyword Planner API to Claude for research — every method, tested.

The google-keyword-planner-mcp repository by ncosentino is a zero-dependency MCP server for the Google Ads Keyword Planner API. You download a single pre-built binary for your OS — no Node.js, no Python, no Go toolchain — drop your five Google Ads credentials in as environment variables, and point Claude Desktop at the binary path in claude_desktop_config.json.

The JSON config block for Claude Desktop looks like this conceptually: you add a keyword-planner entry under mcpServers, set the command to the binary path, and populate the five environment variables (GOOGLE_ADS_DEVELOPER_TOKEN, GOOGLE_ADS_CLIENT_ID, GOOGLE_ADS_CLIENT_SECRET, GOOGLE_ADS_REFRESH_TOKEN, GOOGLE_ADS_CUSTOMER_ID). If you are querying through a Manager account, you also set GOOGLE_ADS_LOGIN_CUSTOMER_ID to your MCC account ID.

The biggest friction is credential procurement. Applying for a Google Ads developer token through a Manager account can take one to three business days, and generating a valid refresh token requires running an OAuth flow manually — a 20-minute process for someone who has not done it before. Once configured, however, the connection is rock-solid, handles all seven API data types, and processes bulk keyword requests reliably. Technical teams who run internal tooling will prefer this to any managed service.

Apify’s GKP MCP Worker is a hosted actor on the Apify platform that connects Claude to Google Keyword Planner data without requiring a Google Ads developer token. You authenticate with a standard Google account (the one associated with any Google Ads or Keyword Planner access), add the Apify MCP connector URL to Claude Desktop, and supply your Apify API token when prompted.

The setup flow inside Claude is: navigate to integrations or MCP settings, paste the Apify connector URL, sign in with your Google account when redirected, then retrieve your Apify API token from Settings in the Apify Console and paste it when prompted. In our test this took eight minutes end-to-end. The tool supports getKeywordIdeas, getHistoricalKeywordData, listAvailableLocations, and getMonthOverMonthAnalysis — four of the seven full-API data types. For most content research workflows that is sufficient, though paid-media forecasting requires the full forecast endpoint which this service does not expose.

The manual CSV upload method requires no API connection at all: you export a keyword ideas or historical metrics report from the Google Keyword Planner interface, upload the file to a Claude Project, and ask Claude to analyze it. This is the method most SEOs start with before discovering how limiting static data is.

For simple tasks — clustering 200 keywords into topic groups, flagging high-volume low-competition opportunities, writing briefs based on existing research — it works fine. The ceiling appears fast: Claude cannot pull additional keywords mid-session, cannot refresh data when you ask about a keyword not in the CSV, and cannot run forecasts. Every new research session requires a fresh export and re-upload. For teams doing keyword research more than twice a week, the time cost of the manual loop exceeds the time it takes to configure a proper MCP connection. See our guide on how to connect Claude to Google and Meta Ads via MCP for the faster alternative.

Windsor MCP positions itself as a data layer between Google Ads and Claude, pulling structured campaign data directly from the Google Ads API. The setup mirrors the Ryze connector in simplicity: connect your Google Ads account at onboard.windsor.ai, select the accounts you want to query, and add the Windsor MCP URL to Claude Desktop.

Where Windsor shines is combining campaign performance data (CPA, ROAS, impression share, search term reports) with Keyword Planner queries in one conversation — asking Claude to compare your current keyword bids against the Planner’s suggested bid ranges, for example. The Keyword Planner surface is less complete than a direct API connection, so pure keyword research teams will hit the ceiling faster. But for paid-search managers who need both lenses, it is a strong option.

DataForSEO is a third-party data API that provides Google search volume metrics, SERP results, and keyword difficulty scores through its own infrastructure. With a single MCP server command in Claude Code — supplying your base64-encoded DataForSEO credentials — Claude gains access to real keyword data in under two minutes.

In community benchmarks, a complete six-phase market research session using DataForSEO MCP in Claude Code — covering keyword ideas, volume validation, competitive analysis, content planning, and brief generation across 8 output files — cost approximately $0.25 in API fees. That is a compelling proposition versus $130/month keyword research subscriptions. The trade-off is that DataForSEO is not the Google Ads API, so you cannot run official Keyword Planner forecasts or access bid estimate data. For content SEO research that does not require media planning data, it is an excellent alternative.

Claude Code with direct Google Ads API keys is the most powerful and most complex option. You configure an MCP server manually in your terminal environment, supply your five Google Ads credentials, and Claude Code gains access to the full API surface — not just Keyword Planner, but the entire Google Ads API if you extend the server.

The setup follows the same credential requirements as the open-source binary, but instead of a pre-built binary you configure the server connection directly in your Claude Code environment. This gives you the ability to write custom tools — for example, a function that pulls keyword ideas, cross-references them against your existing keyword list in a database, and returns only net-new opportunities. For teams with an engineer available, this flexibility is valuable. For everyone else, the 45–60 minute setup and ongoing maintenance burden tips the balance firmly toward managed connectors.

n8n is a visual workflow automation tool that can connect the Google Ads API to Claude through scheduled workflows. A typical setup pulls keyword performance data from Google Ads on a schedule, formats it as structured JSON, and sends it to Claude via the API for summarization or analysis.

This is fundamentally different from an MCP connection: the data flows one-way at the time the workflow runs, rather than Claude being able to query the API on demand mid-conversation. For teams that want a weekly keyword opportunity report delivered to Slack with Claude’s commentary, n8n is excellent. For teams that want to ask Claude “what is the search volume for this keyword I just thought of?” and get an instant answer, n8n is not the right architecture. The community around n8n’s Google Ads and Claude integrations has grown significantly in 2026, with pre-built templates available for common research workflows.

Stormy AI publishes a Google Ads and Claude Code integration playbook built around MCP. Their bridge acts as the translator between Claude Code and the Google Ads API, handling the credential layer so agency users can query across multiple client accounts without reconfiguring credentials each time.

The documented setup installs the official Google Ads MCP Server with your Developer Token, Client ID, and Client Secret from the Google Ads Developer Portal. For agencies managing keyword research across many client accounts simultaneously, the multi-account switching capability is genuinely useful — most other methods require separate configuration per account. The product is newer and less community-tested than Ryze or the open-source binary, so agencies should pilot it on a single account before rolling out broadly.

DataForSEO’s Claude Code integration is worth a separate entry from the MCP version because the Claude Code implementation is distinct: a single terminal command adds the DataForSEO server with your base64-encoded credentials, and the connection persists across all future Claude Code sessions automatically.

In practice, the workflow is remarkable for the cost. A complete market research session — starting from a seed topic, expanding to hundreds of keyword ideas, validating volumes against real Google data, analyzing SERP competition, and generating eight structured research documents including article outlines and social content — costs fractions of a dollar. For freelance SEOs and content strategists who previously paid $99–$200/month for keyword research tools, the economics are hard to ignore. The limitation is that you are working with DataForSEO’s data infrastructure rather than the official Keyword Planner API, which matters if your workflow requires official Google bid estimates for media planning.