MCP integration
Your marketing data, in Claude. In Cursor. In ChatGPT.
Plug Prooflytics into any MCP-enabled AI client through one URL + one token. Ask questions over your real briefings, recommendations, competitor signals, attribution — without copying data around.
Included on Growth ($199/mo) and above. Multi-touch attribution tools require Scale ($449/mo).
Works with every MCP client
One server URL + one Authorization header. Any client that speaks Model Context Protocol can connect.
Six tools your AI can call
Not raw connectors. Curated read-only tools that map to the surfaces marketers actually ask about.
get_latest_briefingMost recent daily / weekly / monthly AI briefing.
list_open_recommendationsPending Action Queue items with the 4-part fact / interpretation / action / consequence.
list_creatives_by_lifecycleAd creatives classified Scaling / Mature / Fatiguing / Dead.
list_competitor_lifecycle_changesRecent competitor ad-library snapshots — themes, ads count, summary.
list_hypothesesHADI hypotheses by status — backlog, in-sprint, validated, learned.
get_attributionScale planMulti-touch attribution credits per channel for any of the 5 models.
Questions you ask in Claude or Cursor
- “What did I miss in the last week, and what should I do today?”→ get_latest_briefing + list_open_recommendations
- “Which Meta ads should I pause, and why?”→ list_creatives_by_lifecycle + list_open_recommendations
- “What are my competitors testing right now that I should consider?”→ list_competitor_lifecycle_changes + list_hypotheses
- “Show me channel contribution under linear vs position-based attribution.”→ get_attribution (×2)
- “What experiments are running, and which ones already produced learnings?”→ list_hypotheses
How it works
- Generate a token in /settings/mcp inside Prooflytics. Plaintext shown once; copy it.
- Paste into your AI client — server URL https://prooflytics.io/api/mcp/<token-id> + Authorization: Bearer <full-token>.
- Restart your client. The Prooflytics tools appear in the available-tools list.
- Ask a question. Your client decides which tools to call; we return JSON; the LLM synthesises a natural-language answer.
- Revoke any time. Tokens are scoped to one workspace + one user; revocation kicks in on the next request.
Frequently asked
What is MCP?▾
Which MCP clients are supported?▾
Which plan includes MCP?▾
What tools are exposed?▾
How is data secured?▾
Try it in your trial
Trial unlocks all features for 14 days, including MCP. Generate a token, paste it into your AI client, ask the first question — under 5 minutes.