ProvenLeap's Claude AI Integration (MCP)

ProvenLeap connects to AI assistants like Claude and ChatGPT through the Model Context Protocol (MCP). Connect ProvenLeap to Claude or ChatGPT and ask about a client's performance in plain language, the assistant reads your live marketing data and can even draft a report, without you exporting anything to a separate AI tool.

What is MCP, and why does it matter for agency reporting?

The Model Context Protocol (MCP) is an open standard from Anthropic that lets an AI assistant like Claude securely connect to external tools and data. ProvenLeap runs an MCP server, so Claude can query your agency's marketing analytics directly. For an agency, that turns reporting from an export-and-paste chore into a conversation: ask a question, get the numbers, act on them.

Learn more about the standard at modelcontextprotocol.io.

How does it work?

The flow is: You → Claude → ProvenLeap MCP → your marketing data. Everything is scoped to your own agency account with strict tenant isolation — Claude can only see your data.

  1. 1Connect ProvenLeap to Claude. Add the ProvenLeap MCP server to Claude. It authenticates against your agency account and only ever sees your own tenant's data.
  2. 2Ask in plain language. Ask Claude about a client — "how did FreshBrew's Google Ads do this month vs last?" Claude picks the right tool and calls ProvenLeap.
  3. 3Get an answer you can act on. Claude returns the numbers with context, and can draft a report — so you go from question to deliverable without leaving the conversation.

What can Claude do with ProvenLeap?

ProvenLeap exposes nine tools to Claude. Eight are read-only analytics; the ninth drafts a report (always as a draft — Claude can never publish to a client on its own).

CapabilityWhat it does
List clientsSee every client in the agency account and their connected platforms.
Client summaryPull a client's headline KPIs for a date range.
Compare two periodsPut this month against last (or any two ranges) for a client.
Metric breakdownBreak a metric down by a dimension — by campaign, by ad set, by channel.
Top performersRank the best-performing campaigns, ads, or clients on a metric.
Search metricsFind which metrics are available for a client across their platforms.
List reportsSee the reports that already exist for a client.
Task statusCheck on a long-running job.
Create reportDraft a new report (created as a draft — never auto-published to a client).

Example ways agencies use it

  • "How did this client's Google Ads perform this month versus last?"
  • "Which campaigns had the best ROAS across all my clients last quarter?"
  • "Draft a monthly performance report for this client."
  • "What metrics do I have available for this client's LinkedIn Ads?"

Getting started

When you connect ProvenLeap to Claude, you add the ProvenLeap MCP server and authorize it against your agency account. From then on, Claude can answer questions about your clients and draft reports. The connection is read-oriented and tenant-isolated by design.