Funnel MCP brings your marketing data, field definitions and workspace context into the AI products you already use.

The Model Context Protocol (MCP) is an open standard for connecting apps to external tools and data. Many apps support MCP, so you can use Funnel MCP in the client that fits your workflow.

Funnel MCP authenticates with OAuth 2.0. The tools are read-only: they read data and definitions from your account. They do not create or change data sources, fields, dashboards or other configuration in Funnel. Funnel MCP is available for all plans and doesn't consume flexpoints.

Funnel MCP is a good fit if your team wants to work with Funnel data from an MCP-compatible client, for example an AI assistant or IDE.

Compatible MCP clients

Funnel MCP works with the following MCP clients:

Anthropic Claude (listed in the Claude connector directory): See Connect Funnel MCP to Claude for setup.
ChatGPT (listed in the ChatGPT plugin directory): See Connect Funnel MCP to ChatGPT for setup.
Claude Code: See Connect Funnel MCP to Claude Code for setup.
OpenAI Codex
Cursor
Gemini CLI and Google Antigravity
GitHub Copilot in VS Code
Lovable
Perplexity
Other MCP-compatible clients (manual config)

Copy connection details from Funnel AI > MCP. See Connect Funnel MCP from Funnel AI.

Prerequisites

Funnel account with access to at least one workspace.
Connect at least one data source in that workspace so MCP queries return real results.

Authentication and regions

Funnel MCP requires OAuth 2.0 authentication.
Use the account, from your Funnel subscription, that has access to your workspaces.
Funnel runs MCP on regional infrastructure. Use the MCP server URL or region based on where your data resides.

Where does your data reside?	Funnel subscription URL	Funnel MCP server URL
Global (default)	`https://app.funnel.io/`	`https://mcp.ai.funnel.io/mcp`
EU	`https://app.eu.funnel.io/`	`https://mcp.eu.ai.funnel.io/mcp`

Note: Do not guess MCP server URLs. Use the values from the official Funnel connection experience.

Available tools

Funnel MCP exposes read-only tools that your MCP client calls on your behalf. You describe the outcome in natural language. The client selects the tools.

Workspace preparation

Tool	Description
`list_workspaces`	Lists Funnel workspaces you can access, grouped by subscription. Use the returned workspace IDs in later tools.
`load_workspace`	Loads connected platforms, field counts, reporting currency and data source health for one workspace. Required before you search fields or query data.
`get_workspace_context`	Returns administrator-authored conventions, business rules and domain terminology for the workspace. Call once after `load_workspace`.

load_workspace can report sources that are unauthorized, delayed or invalid. Ask the client to summarize source health when a query returns less data than you expect.

Field discovery

Tool	Description
`search_fields`	Finds dimensions, measures and metrics in the loaded workspace.
`get_dimension_values`	Returns actual values for a dimension in a date range, for example campaign names or countries.

search_fields has three modes:

Mode	When to use it
semantic	Preferred for discovery. Ask in natural language (for example spend on Facebook or cross-platform performance metrics).
browse	Explore a known platform, category or group, or list data sources and their IDs.
search	Exact field name or label only.

Some custom metrics appear with partial availability. Results are correct for the platforms that report the underlying data and can be empty for platforms that do not.

Pull and analyze data

Tool	Description
`query_data`	Queries advertising data from connected platforms and returns a table for further analysis.
`prepare_data`	Loads raw, row-level data without aggregation.

query_data modes:

Mode	When to use it
query	Fetch fresh data by dimensions and measures for a date range.
compare_channels	Compare metrics across platforms using cross-platform fields only.
aggregate	Re-slice a table you already loaded, without fetching again.
compute_metric	Calculate a ratio such as CPC, CTR or ROAS on a loaded table.
restore	Reload a result that was evicted from session memory.

How MCP clients typically use these tools

Select a workspace: Run list_workspaces, then run load_workspace for the workspace you need.
Load workspace context: Run get_workspace_context once after load_workspace so answers use your team's conventions and terminology.
Explore fields and dimension values: Run search_fields to find fields to query. Run get_dimension_values when you need values for a specific dimension.
Pull and analyze data: Use query_data for aggregated analysis and cross-channel comparisons. Use prepare_data if you need raw, row-level data for custom slicing.

Note: Tool order is determined by your MCP client. Describe the outcome you want in natural language. The client selects and runs the tools.

If query_data returns a truncated table, ask the client to re-slice the result (for example a coarser breakdown, a filter or a shorter date range) instead of raising the row limit. query_data returns up to 500 rows by default. The full result stays on the server.

Connect other MCP clients

For clients that are not listed in a marketplace, paste the regional Server URL into the client's MCP settings and complete OAuth when prompted.

Cursor example (global region):

{
  "mcpServers": {
    "funnel": {
      "url": "https://mcp.ai.funnel.io/mcp"
    }
  }
}

VS Code Copilot (agent mode) example:

{
  "servers": {
    "funnel": {
      "type": "http",
      "url": "https://mcp.ai.funnel.io/mcp"
    }
  }
}

Replace https://mcp.ai.funnel.io/mcp with https://mcp.eu.ai.funnel.io/mcp when your Funnel data resides in the EU.

If the client requires a static OAuth client ID (for example some Copilot setups), copy OAuth Client ID from Funnel AI > MCP > Other MCP clients.

Example use cases

Use these prompts as starting points. Replace workspace names, date ranges and field names to match your setup.

Cross-channel marketing performance

Get a trusted cross-channel view of what is driving performance, from high-level channel comparison down to specific campaigns, ads, creatives and keywords.

Show me the trends for spend, impressions, clicks, CPC, and revenue. Break it down by traffic source for the last six months.
Compare attribution models between the platforms and Google Analytics. How do they compare for the last month?
Give me an executive weekly report of my KPIs with sessions, transactions, revenue, AOV, and CVR.

Pacing towards budget and targets

Stay on top of how spend is pacing against budget and how KPIs such as revenue are performing against targets.

Show month-to-date spend by channel against monthly budget. Flag any channel pacing more than 10% over.
Which traffic sources will overspend their budget by month-end if current pace continues?
At current pace, will we hit our Q2 revenue target? Show the gap and where it is coming from. What should I do about it?

Data exploration and deep-dives

Drill into a specific channel, campaign, product or market. Surface anomalies, find underperformers and get recommendations for further analysis or action.

CPA on Google Ads spiked last week. Which campaigns and keywords are driving the change?
Break down conversions on Facebook by campaign and creative for the last 30 days.
Drill into our top five campaigns by transaction volume. Which paid channels are feeding them? What is the split between organic and paid traffic?

Security and support

Read the Funnel privacy notice for privacy practices, including AI-related topics described there.
For support guidelines, the AI support assistant and how to contact the team, read Funnel support 101.

Quick start guide: Using Funnel MCP