Google Ads Intent MCP README
The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Google Ads Intent MCP listing page.
Google Ads Intent MCP
Dry-run-first Google Ads search-term intent analyzer + negative-keyword MCP for agents.
Look at search terms safely before spending budget.
mcp-name: io.github.davidmosiah/google-ads-intent-mcp
⭐ If this agent-first tool helps your workflow, please star the repo. Stars make this tooling easier for other builders to discover and help Delx keep shipping open infrastructure.
🧱 Part of the Delx agent stack — 15 open-source MCP servers across body, reach and coordination.
Dry-run-first Google Ads search-term intent analyzer for agents. It helps Codex, Claude, Cursor, Hermes, OpenClaw and other MCP clients classify search terms, protect buyer intent and draft negative-keyword plans from CSV exports before any live account change.
Use it when an agent needs to reduce wasted spend without accidentally excluding buyer-intent queries.
HTTP (v2 stateless)
Not shipped. Optional FastMCP extra is stdio-only (same skip as delx-agent-utilities: no /health without extra Starlette routes). Dry-run analysis stays request-stateless over stdio.
Why It Exists
Google Ads cleanup is risky when agents act directly on accounts. This package makes the safe path the default:
- analyze exported search-term CSVs locally
- classify waste, buyer, research and competitor intent
- draft negative-keyword plans without applying them
- expose
manifest,connection_statusandprivacy_auditbefore action tools - keep live mutation out of v0.1
Install
With MCP support:
Published on PyPI: google-ads-intent-mcp. Release automation uses PyPI Trusted Publishing, so GitHub Actions can publish future versions without long-lived PyPI tokens. See docs/pypi-publishing.md.
CLI
Intent classification
The classifier is a deterministic, dependency-free heuristic with broad,
cross-vertical signal coverage (ecommerce, B2B/SaaS, local services, health,
finance, education and more) — not just gaming traffic. It sorts each search
term into waste, buyer, research or competitor intent and protects
converting queries from being flagged as negatives.
An optional LLM/embeddings-backed refinement path is available and is off by default. It requires no extra dependencies or API keys for normal use, and always falls back to the heuristic when no backend is configured:
To actually call a backend, set OPENAI_API_KEY (and optionally
GOOGLE_ADS_INTENT_LLM_MODEL, default gpt-4o-mini) and install the openai
package. Without those, --llm is a no-op that keeps the heuristic result.
Each classification reports which path produced it via a source
(heuristic or llm) field.
MCP
Hermes-style config:
Recommended first calls:
google_ads_connection_statusgoogle_ads_privacy_auditgoogle_ads_analyze_search_termsgoogle_ads_build_negative_plan
Agent Surfaces
| Tool | Purpose |
|---|---|
google_ads_agent_manifest | Install/runtime guidance for agent clients |
google_ads_connection_status | CSV/API readiness without credentials |
google_ads_privacy_audit | Dry-run, account and export boundaries |
google_ads_classify_search_term | Single-query intent classification |
google_ads_analyze_search_terms | Batch CSV-style analysis |
google_ads_build_negative_plan | Dry-run negative keyword plan |
Copy-Paste Agent Prompt
CSV Format
The parser accepts common exported columns such as:
search_term,Search term,Querycost,Cost,cost_microsclicks,Clicksconversions,Conversions,Conv.impressions,Impr.,Impressions
Safety Model
- CSV analysis is local.
- Negative plans are dry-run only.
- Buyer/conversion terms are protected from automatic exclusion.
- OAuth tokens, developer tokens and account identifiers should stay in local environment/config files.
Development
📧 Contact & Support
- 📨 support@delx.ai — general questions, integration help, partnerships
- 🐛 Bug reports / feature requests — GitHub Issues
- 🐦 Updates — @delx369 on X
- 🌐 Site — wellness.delx.ai