データハブ
AI Analysis (ChatGPT & Claude)
Connect AdStage to your AI assistant once, and it queries Data Hub connectors (ad platform cost, AdStage attribution, blended data) to answer your questions. The connection is read-only and limited to the organizations you approve.
- Server URL:
https://api.adstage.app/mcp(copy it from Data Hub > AI Analysis) - Authentication: sign in with your own AdStage account and approve (OAuth). No keys to paste.
1. Connect
| Assistant | How |
|---|---|
| Claude (claude.ai, Desktop) | Settings > Customize > Connectors > Add > Add custom connector → name AdStage, server URL → Add → Connect → sign in, choose organizations, approve. Team/Enterprise: an Owner adds it under Organization settings > Connectors first |
| ChatGPT | Settings > Integrations > Plugins > Add > Add MCP server → name AdStage, type Streamable HTTP, server URL, authentication OAuth → connect, sign in, approve. Older UI: Settings > Connectors > Advanced, turn on Developer mode, then Create |
| Claude Code | Run claude mcp add --transport http adstage https://api.adstage.app/mcp, then /mcp in a session |
| Cursor, Gemini CLI, others | Add a remote MCP server (Streamable HTTP), e.g. { "mcpServers": { "adstage": { "url": "https://api.adstage.app/mcp" } } } |
Menu details by plan, connecting several organizations, revoking and expiry are covered in AI Assistants.
2. Data Hub tools
Once connected, the assistant also sees these tools and picks them as needed.
| Tool | What it does |
|---|---|
list_data_connectors | Connected sources and their status (connected, check sync, available, coming soon) |
get_connector_fields | A connector's fields, accounts, date presets and filter operators |
get_connector_data | windsor.ai-style query — any fields, range and and/or filters, up to 500 rows per call |
list_data_destinations | How to send data to Sheets, Looker Studio, Power BI, Excel, Python, webhooks |
get_connector_connect_url | Dashboard link to connect a source (connecting happens in the browser) |
The existing tools (query_attribution_report, get_funnel_report, query_retention_report, diagnose_tracking_setup …) remain available.
3. Try asking
- Compare Google Ads, Meta and TikTok for the last 30 days: spend, clicks, CPM, installs and ROAS. Which one should get more budget?
- Show me Meta ad sets with spend over 100,000 and no purchases last week.
- Weekly trend of spend vs AdStage revenue across all channels for the last 90 days, as a table.
- Which Naver Search Ads keywords had the highest CPC this month, and how many purchases did they drive?
- Compare cost per install between Instagram and Facebook placements.
- Spend looks high but installs are low. Is anything wrong with our tracking setup?
4. Tips
- Name the organization. If you connected several, include its name in the question (e.g. "TripTok spend by channel last week").
- Be explicit about dates. "Last month" or "last 4 weeks" are resolved in the organization timezone; presets end yesterday.
- Large tables (thousands of rows) belong in the Data API or a sheet. Tools return up to 500 rows per call.
- If numbers differ from the dashboard, check the range, filters and attribution window. Same conditions give the same values.
- See what the assistant queried under Settings > AI Assistants > Call history.

