Works with
Works in any MCP client.
Pathbound is one remote MCP server, so the recipe runs wherever your agent does. Pick the product your team already uses and follow its setup.
Don’t see yours? Any MCP client connects the same way.
For a founder or CS lead running retention without an analyst, the job is: every Monday, know which accounts dropped, why, and which three to call—in a digest short enough to read before standup.
The plan was a standing Monday review: what dropped, why, who calls. In practice the “what” means exporting activity by account, the “why” means opening support threads one at a time, and by the time both are assembled it’s Wednesday and the review is a skim. So the ritual quietly dies. Churn gets diagnosed once a quarter, in hindsight, at the renewal that doesn’t close.
One profile per account already holds the joins—activity from Events, conversations from every channel, CRM state. The diagnostic is four calls to Pathbound’s MCP server, and the expensive-looking part, scanning every account, is the free part.
aggregate_data Two server-side aggregations—this week’s events by account, last week’s beside them—and every mover surfaces: the accounts that dropped, the ones that went silent outright, the ones that spiked. The whole book gets scanned without pulling a single profile.
batch_get_details Up to 25 accounts per call turns a column of moved IDs into a working list: company, size, industry, CRM state. The paying account that dropped by half reads differently from a free workspace that went dark, and this is the step that tells them apart.
get_company_conversations The drop usually has a reason sitting in the threads: a failure alert nobody answered, an onboarding email that stalled, a support ticket that ended cold. Rolling up every contact’s conversations at the account puts the reason next to the number.
create_company_note Each flagged account gets a note—what dropped, the likely why, the next step—and the agent hands back the week’s digest with the three accounts to call first. Notes live on the unified profile every connected agent reads; syncing notes into two-way CRMs is rolling out.
Same four calls every Monday. In Cowork, save it once and it becomes the team’s standing task.
First, connect Pathbound to your agent—about five minutes, in whichever client your team runs. Then paste this prompt. This is the weekly retention ritual, automated: save it as a recurring Monday task in Claude Cowork (or any scheduler that speaks MCP) and read the digest over coffee.
Run our weekly churn diagnostic. Compare every account’s activity this week against last week and list the accounts that dropped the most, went fully quiet, or spiked. For each account that cooled, pull its details and read its recent conversations—support threads, failure alerts, email that went cold—and give me a one-line reason why. Leave a note on each flagged account with the read and the suggested next step, then give me the digest: what moved, why, and the three accounts I should call today.
The Monday review happens, because it costs a prompt instead of a half-day of exports. Drops get a reason attached while the account can still be saved—the failure alert answered, the stalled setup rescued. We ran this flow on a live workspace while writing this page: the week-over-week scan surfaced an account whose activity had fallen off a cliff, and the conversation read explained it—unanswered failure alerts and a setup thread that had gone quiet.
Pathbound is one remote MCP server, so the recipe runs wherever your agent does. Pick the product your team already uses and follow its setup.
Don’t see yours? Any MCP client connects the same way.
Yes—this recipe is a weekly churn analysis an agent runs end to end: week-over-week activity by account, reasons pulled from the actual conversations, a digest with the accounts to call. Pathbound is the customer data MCP it runs against; the agent—Claude, Cowork, n8n, or your own loop—stays yours.
Yes—this is the churn workflow Cowork fits best: a repeatable multi-step task with a document at the end. Save the prompt as a recurring Cowork task and it runs Monday mornings, reading Pathbound over MCP and returning the digest. Setup is on the use Pathbound with Claude Cowork page.
The quiet-account recipe is the watchlist: a scan that ranks accounts by how far they cooled, tuned to run ahead of renewals. This is the ritual on top—a weekly digest that also reads why each account moved and writes the verdict down. Run the watch monthly and the diagnostic weekly; they share the same profiles underneath.
From the account’s own record: support threads from Intercom or Zendesk, email that went cold, failure alerts, meeting notes—rolled up across every contact at the company with get_company_conversations. The agent quotes what’s there. When the record is thin, the honest answer is “activity dropped, no signal why,” which is itself a reason to call.
Pathbound tracks activity, conversations, and CRM state—not billing. If your CRM carries plan or deal fields, the agent reads them with the account details; the revenue math stays in your billing tool. The diagnostic tells you who cooled and why, which is the part billing can’t see.
A note per flagged account—what dropped, the likely reason, next step—written with create_company_note, plus the digest the agent hands back in chat. Notes live on the unified profile every connected agent reads; syncing notes into two-way CRMs like HubSpot is rolling out.
Sign up, connect one CRM or drop the events snippet, and point your MCP client at
mcp.pathbound.ai/mcp.