ConvertFlow MCP Server

The ConvertFlow MCP server turns your funnel reports into tools an AI agent can call. It's a remote server: Streamable HTTP, stateless, and hosted at https://mcp.convertflow.co/mcp. There's nothing to install and nothing to run locally. Connect a client, sign in through your browser, and ask questions about your campaigns in plain language.

Connect

Claude Code

claude mcp add --transport http convertflow https://mcp.convertflow.co/mcp

Run any ConvertFlow tool, and Claude Code opens a browser so you can authorize it.

Claude.ai and Claude Desktop

Go to Settings → Connectors → Add custom connector and paste https://mcp.convertflow.co/mcp. As you add it, Claude prompts you to sign in and pick the websites you want to share.

Cursor

Add the server to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "convertflow": {
      "url": "https://mcp.convertflow.co/mcp"
    }
  }
}

Cursor opens the sign-in prompt the first time you use the server.

Any other Streamable HTTP client

{
  "mcpServers": {
    "convertflow": {
      "type": "http",
      "url": "https://mcp.convertflow.co/mcp"
    }
  }
}

Your client opens ConvertFlow in a browser. Sign in, choose which websites to share, and approve. The token your client receives carries the reports scope and is bound to the websites you picked.

How authorization works

The MCP server is a pure resource server, so it never issues or stores tokens. On an unauthenticated request it returns 401 and points your client at /.well-known/oauth-protected-resource, which names ConvertFlow as the authorization server. Your client then registers itself dynamically, runs OAuth 2.1 authorization code + PKCE against ConvertFlow, and sends the resulting token as Authorization: Bearer <token>. Tokens are website-bound. If a token expires or is revoked, the next call returns 401 and your client re-authorizes.

Available tools

Every tool maps to a v1 Reports endpoint. Start broad with get_website_report, then drill into campaigns, variants, steps, and elements using the ids that each response returns.

list_websites

List the websites this token grants access to. Use the returned ids as website_id on the other tools to query or compare specific websites.

No parameters.

get_website_report

Website-wide funnel performance (views, conversions, completions, subscribes, orders, revenue) for the selected window. Returns one total per metric; add group_by=day (or week/month/quarter/year) for a per-bucket breakdown, over a window of up to 92 days. Start here, then drill into campaigns.

Parameter	Type	Description
`group_by`	string	Adds a timeseries block: one point per bucket across the selected window, each with date, views, conversions, completions, subscribes, submissions, orders and revenue for this report's subject. Buckets with no activity come back as zeroes, so the series is complete. Use group_by=day for a daily breakdown instead of calling the tool once per day. On get_website_report the window must cover at most 92 days (the other reports take any range up to 400 buckets); a wider one is rejected — shorten it, or pull the longer series from a campaign, variant, step or element report. one of: day, week, month, quarter, year
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

get_website_insights

Latest pre-generated weekly insights snapshot (one fixed week, written by a background job). Use it for the narrative summary only — for any specific window or trend use get_website_report, which reads live per-day data. Returns 404 if no snapshot has been generated yet.

Parameter	Type	Description
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.

list_campaigns

List campaigns (CTAs) with their funnel metrics for the selected window. Returns the campaign ids used by get_campaign_report. Supports pagination, sorting and filtering by type/status.

Parameter	Type	Description
`page`	integer	Page number (default 1).
`per_page`	integer	Results per page (default 25, max 50).
`sort`	string	Sort field. one of: created_at, published_at, views, conversions, completions, conversion_rate, completion_rate, revenue, aov
`cta_type`	string	Filter by campaign format.
`status`	string	Filter by status. one of: live, archived
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

get_campaign_report

Single campaign performance with per-variant breakdown. Campaign ids come from list_campaigns; the variant ids in the response feed get_variant_report.

Parameter	Type	Description
`id` *	integer	The campaign id.
`group_by`	string	Adds a timeseries block: one point per bucket across the selected window, each with date, views, conversions, completions, subscribes, submissions, orders and revenue for this report's subject. Buckets with no activity come back as zeroes, so the series is complete. Use group_by=day for a daily breakdown instead of calling the tool once per day. On get_website_report the window must cover at most 92 days (the other reports take any range up to 400 buckets); a wider one is rejected — shorten it, or pull the longer series from a campaign, variant, step or element report. one of: day, week, month, quarter, year
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

list_experiments

Active A/B experiments — campaigns with live traffic-split variants and their metrics.

Parameter	Type	Description
`limit`	integer	Max experiments returned (default 25, max 50).
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

get_variant_report

Variant funnel performance with per-step breakdown. Variant ids come from get_campaign_report; the step ids in the response feed get_step_report.

Parameter	Type	Description
`id` *	integer	The variant id.
`group_by`	string	Adds a timeseries block: one point per bucket across the selected window, each with date, views, conversions, completions, subscribes, submissions, orders and revenue for this report's subject. Buckets with no activity come back as zeroes, so the series is complete. Use group_by=day for a daily breakdown instead of calling the tool once per day. On get_website_report the window must cover at most 92 days (the other reports take any range up to 400 buckets); a wider one is rejected — shorten it, or pull the longer series from a campaign, variant, step or element report. one of: day, week, month, quarter, year
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

get_step_report

Step funnel performance with per-element breakdown. Step ids come from get_variant_report; the element ids in the response feed get_element_report.

Parameter	Type	Description
`id` *	integer	The step id.
`group_by`	string	Adds a timeseries block: one point per bucket across the selected window, each with date, views, conversions, completions, subscribes, submissions, orders and revenue for this report's subject. Buckets with no activity come back as zeroes, so the series is complete. Use group_by=day for a daily breakdown instead of calling the tool once per day. On get_website_report the window must cover at most 92 days (the other reports take any range up to 400 buckets); a wider one is rejected — shorten it, or pull the longer series from a campaign, variant, step or element report. one of: day, week, month, quarter, year
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

get_element_report

Element performance with its field and product breakdown — form field values, quiz answer counts and per-product orders/revenue. Element ids come from get_step_report.

Parameter	Type	Description
`id` *	integer	The element id.
`group_by`	string	Adds a timeseries block: one point per bucket across the selected window, each with date, views, conversions, completions, subscribes, submissions, orders and revenue for this report's subject. Buckets with no activity come back as zeroes, so the series is complete. Use group_by=day for a daily breakdown instead of calling the tool once per day. On get_website_report the window must cover at most 92 days (the other reports take any range up to 400 buckets); a wider one is rejected — shorten it, or pull the longer series from a campaign, variant, step or element report. one of: day, week, month, quarter, year
`website_id`	integer	Website to query. Optional for single-website tokens; required when the token grants multiple websites — call list_websites first.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.
`count_type`	string	distinct = unique visitors via HyperLogLog (default); total = raw event counts. Orders/revenue are always summed. Distinct counts deduplicate across the whole window, so they do not add up across sub-periods — use total when separate calls or timeseries points must reconcile with a longer range. one of: distinct, total
`previous_period`	boolean	When true, each metrics block includes the previous period of equal length and percent changes.

get_account_performance

Cross-site funnel performance for every website bound to this token: per-site rows plus an aggregate. Omit website_id to include all bound sites; pass website_id or website_ids to zoom in. Does not require website_id on multi-website tokens.

Parameter	Type	Description
`website_id`	integer	Optional. Zoom to one website bound to the token. Omit to aggregate every bound website. Unbound ids return 404.
`website_ids`	array<integer>	Optional subset of bound website ids. Any id outside the token binding returns 404.
`timeframe`	string	Reporting window (default last_30_days). Use custom with range_start/range_end for any exact range, including a single day (same date on both). Windows are UTC. one of: today, this_week, this_month, this_quarter, this_year, last_24_hours, last_7_days, last_30_days, last_90_days, last_365_days, all_time, insights_week, custom
`range_start`	string	ISO 8601 date (YYYY-MM-DD), inclusive from 00:00:00 UTC. Required when timeframe=custom.
`range_end`	string	ISO 8601 date (YYYY-MM-DD), inclusive through 23:59:59 UTC. Required when timeframe=custom.

Notes & limits

Reports only, for now. The server is read-only, so it can't edit campaigns or write contacts.
Multiple websites need website_id. A token bound to one website doesn't need it. A token bound to several does. Call list_websites to get the ids. The exception is get_account_performance, which aggregates every bound website.
Freshness matches the app. Tools read the same per-day metrics as your in-app reports. There's one exception: get_website_insights returns a pre-generated weekly snapshot, not live data.