Skip to content
Helferlain

AI integration

Google Ads in Claude, ChatGPT and Cursor.
Via MCP, with approval.

Helferlain is an MCP server. MCP is an open standard that lets AI assistants use external tools. Connect Helferlain to your assistant and ask your Google Ads account directly, with a dry run before every change.

Request accessSetup guide for Claude
  • 46 tools
  • Dry run by default
  • Counts towards your plan quota
Illustration

Run an audit of my account and name the three most important priorities. Start with the tracking check.

Helferlain tools

  • google_ads_audit_conversion_tracking
  • google_ads_run_full_audit
  • google_ads_find_optimization_priorities

Answer with priorities

Changes only as a dry run

Illustration based on the documented layout of the app, without real account data.

Connect

In your assistant in three steps

The MCP server address is the same for every assistant. The exact steps are in the setup guides.

MCP addresshttps://mcp.helferlain.app/mcp

  1. Add the address

    In your assistant, create a custom connector or MCP server and paste the address above.

  2. Choose a workspace

    A Helferlain window opens. Choose your workspace there and confirm with "Approve".

  3. Ask questions

    Name the account, the period and the goal. The numbers come from Helferlain; the assistant writes the answer.

Claude Code

claude mcp add --transport http helferlain https://mcp.helferlain.app/mcp

Tools

46 tools, one of which writes

45 tools read: audit, priorities, wasted spend, tracking, bids, geo, brand, comparisons, anomalies, GA4, Search Console and expert knowledge on Google Ads topics. Changes go only through google_ads_mutate_with_approval, and only with an approval token.

The Bigin tools exist; the Bigin connector in the app has been announced.

Show all 46 tools

Audit and priorities

  • google_ads_run_full_audit
  • google_ads_audit_full_export
  • google_ads_audit_account_health
  • google_ads_audit_cross_source
  • google_ads_find_optimization_priorities
  • google_ads_recommend_optimizations_priority
  • google_ads_audit_history
  • google_ads_helferlain_consultation
  • google_ads_get_account_profile

Wasted spend and keywords

  • google_ads_audit_wasted_spend_deep
  • google_ads_analyze_wasted_spend_ngram
  • google_ads_analyze_keyword_duplication
  • google_ads_detect_language_keyword_gaps

Tracking and measurement

  • google_ads_audit_tracking_complete
  • google_ads_audit_conversion_tracking
  • google_ads_audit_measurement_readiness
  • google_ads_audit_oci_status

Bids, budget, geo, audiences

  • google_ads_audit_bidding_and_budget
  • google_ads_audit_geo_ad_scheduling
  • google_ads_audit_audience_targeting
  • google_ads_geo_target_lookup

Brand, competition, landing pages

  • google_ads_analyze_brand_cannibalization
  • google_ads_competitor_pressure_report
  • google_ads_audit_landing_page_routing

Comparisons, anomalies, changes

  • google_ads_compare_periods
  • google_ads_detect_anomalies
  • google_ads_list_change_events
  • google_ads_list_change_status

Accounts, campaigns, queries

  • google_ads_list_accessible_customers
  • google_ads_list_campaigns
  • google_ads_get_campaign_performance
  • google_ads_search_stream_gaql
  • helferlain_list_workspaces

Google Analytics 4

  • ga4_run_report
  • ga4_compare_periods
  • ga4_list_properties

Search Console

  • gsc_search_analytics_query
  • gsc_list_sites
  • gsc_run_seo_audit

CRM (Bigin) and Tag Manager

  • bigin_list_deal_fields
  • bigin_get_lead_attribution
  • bigin_analyze_revenue_cohorts
  • gtm_list_containers

Google Ads knowledge

  • google_ads_get_ppc_playbook_for_topic
  • google_ads_get_industry_benchmarks

Writing, only with an approval token

  • google_ads_mutate_with_approval

As of: tool list served live by the MCP server, register dated 18 October 2026.

Writing

Reading, yes.
Writing only with a token.

By default, the change tool only checks a change as a dry run at Google and changes nothing. A real change needs an approval token that is valid for exactly this change, for only a few minutes and only once. The app does not currently show such a token for assistants, so real changes run through the app or the agency console.

What you see when a change is rejected

  • cpc_cap_exceeded

    A bid is above €10 or cannot be read.

  • budget_cap_exceeded

    The budget is above the cap.

  • mutation_shape_forbidden

    The change has a form that is not allowed, for example a new campaign that is not paused.

  • r2_approval_required

    This change is missing its approval.

All safety limits

Good questions

Seven questions to start with

Name the account, the period and the goal. The more specific you are, the better the answer.

  1. 1. Account audit

    “Run an audit of my account and name the three most important priorities. Start with the tracking check.”

  2. 2. Wasted spend

    “Show me search terms from the last 30 days with cost and no conversions. Which would you suggest as negative keywords?”

  3. 3. Check bids

    “Check bid strategies and budgets. Is a budget holding back campaigns that perform well?”

  4. 4. Anomalies

    “Were there any unusual outliers in cost, clicks or conversions in the last 14 days?”

  5. 5. Comparison

    “Compare the last 30 days with the 30 days before and explain the biggest differences.”

  6. 6. Changes

    “What was changed in the account in the last two weeks?”

  7. 7. Plan a new campaign

    “Plan a Search campaign for the Middle Franconia region with a €20 daily budget. Look up the location IDs, propose ad groups, keywords and ads, and check whether the areas overlap with an active campaign.”

How a plan becomes a campaign proposal

Quota

Every call counts

The connection counts towards the tool call quota of your plan. One question can trigger several calls.

Pricing and limits

  • Starter

    200

    tool calls per month

  • Professional

    1,000

    tool calls per month

  • Agency

    5,000

    tool calls per month

To be clear

What does not work in your assistant

  • No real changes directly from your assistant, as long as the app does not issue an approval token for it.
  • Protected accounts stay locked, even for dry runs.
  • The assistant writes the answer; the numbers come from Helferlain. Double-check important statements in the report.

Questions

Questions about the AI integration

More in the help center (in German)

Do I need my own AI key?
No. You can start without one. If you prefer, you can add your own Anthropic or OpenAI key in the settings; workspace owners and admins manage this.
Which Claude or ChatGPT plans work?
Your assistant must allow custom connectors or MCP servers. In Claude, the number of custom connectors is limited depending on the plan; in ChatGPT, availability as an app or connector depends on the plan.
What happens to my data at Anthropic or OpenAI?
If you use Helferlain in Claude or ChatGPT, the privacy terms of these providers also apply to the conversations. What Helferlain itself reads and stores is described on the security and data protection page.
Can the assistant change my account?
Not today. Without an approval token, every change runs as a dry run. A real change would need a token that works once and is time-limited, and the app does not currently show such a token for assistants. Real changes therefore run through the app or the agency console, and fixed limits such as bids capped at €10 and new campaigns only paused always apply.

Ask your account from your assistant

Request access, connect your ad account in Helferlain, then add the MCP address in your assistant.

Request accessGoogle Ads audit
Stock photo: Open laptop on a wooden table between a picture frame and a flowering cactus
Stock photo