Agents see read-only churn context.
ChurnWin gives AI agents a safe, bounded view of the churn data they need to answer questions, draft recommendations, and connect feedback to product work — without giving them the keys to modify billing or message customers.
What agents can see
Agent keys expose the data needed for analysis and automation. They are intentionally scoped to read-only resources, so agents can reason about churn without taking account actions on their own.
Churn metrics and trend
Churn rate, MRR, retention, subscriber counts, and daily churn movement for a selected period.
At-risk customers
Subscribers ranked by AI risk score, plan, MRR, and the warning signals that make them worth reviewing.
Cancellation feedback
Verbatim answers from churned customers plus reason categories and submitted timestamps.
Churn reasons
Aggregated reason breakdowns so an agent can compare themes before recommending fixes.
Customers and businesses
Paginated customer rosters and the connected Stripe accounts the key is allowed to read.
AI action plan
ChurnWin's prioritised retention recommendations, rationale, and effort notes.
Same context, two safe connection paths.
Use MCP when you want an AI client to discover tools automatically. Use REST when you want standard JSON endpoints for internal agents, scripts, or reporting jobs.
Run a failed-payment recovery brief
Give the agent an explicit evidence contract instead of asking for generic retention advice. This example uses fields already exposed by ChurnWin's read-only metrics and recovery tools; it does not invent a recovery result or authorize an account action.
Copyable prompt
Review the last 30 days of failed-payment churn. Use get_metrics, list_recent_churn, and get_action_plan. Report payment_failures, recovered_count, recovered_mrr, and at_risk_mrr. Separate observed values from recommendations. Do not contact customers or change billing.
What the agent should return
- 1
Recovery snapshot
The observed failure count, recovered subscriptions and MRR, and revenue still at risk, with the requested time window stated.
- 2
Evidence trail
Which recent failed-payment events and action-plan evidence support each finding, without exposing the API key or copying customer data into a public channel.
- 3
Human-reviewed next step
A short ranked recommendation for a person to review before any billing, customer-contact, or production change.
Let agents analyse; keep humans in control.
The agent surface is designed for summaries, comparisons, triage, and suggested next actions — not for unsupervised account changes.
Ask Claude what changed in churn this month and which feedback quotes prove it.
Have an internal agent draft weekly retention priorities from metrics, reasons, and at-risk accounts.
Let a roadmap assistant cross-reference cancellation feedback with planned product work.
Pipe ChurnWin's read-only endpoints into a Slack or Linear workflow without giving it write powers.
Safe by default
Agent keys can read churn context, but cannot change Stripe, email customers, or delete data. Humans keep the dashboard, settings, billing, and customer-contact controls.