Account health without the gut feel
Nobody needs to read every Slack thread to know if an account is healthy — that's an inference problem, not a willpower problem. These blueprints read the scatter, Slack, usage, audit, and return the same structured health read every time.
SlackWhat breaks when you do this by hand
Account health lives in someone's head, read fresh out of Slack threads every time.
Adoption and reliability data sits in tables no one queries by hand — so health calls run on gut feel.
Tenant reviews mean rebuilding the same deck from scratch, every quarter.
What you get instead
One structured health read — usage, audit, and error data, reconciled automatically
Scheduled reporting instead of a person scrambling before every QBR
The same signal, read the same way, across every account
This is the whole thing
The opening of Customer Health Report — the system prompt, its typed parameters, and the tools it's allowed to reach, all declared up front. No canvas, no hidden nodes. 221 lines of source you can review in a pull request.
Read all 221 lines →customer-analytics.fml
system("You are a CS and DevOps analyst generating a customer health report.") parameter("tenant_id", type=string, title="Tenant ID") parameter("customer_name", type=string, title="Customer Name") require mcp BigQuery components { schema("ServerStats") { mcp_name: string status: string total_tool_calls: int total_errors: int error_rate: float active_users: int distinct_tools_called: int active_days: int first_call: string last_call: string days_since_last_call: int connected_but_never_called: bool }
3 blueprints you can run today
Every one is a typed FML plan. Open it, read the source, and run it — nothing here is a mockup.
Customer Health Report
Health calls run on gut feel because the data sits in warehouse tables nobody queries by hand.
Read the blueprint
Customer Slack Signals
Account health lives in scattered Slack threads. No one has a repeatable way to read it — so nobody really does.
Read the blueprint
Tenant Usage Report
A dozen ad-hoc warehouse queries, rebuilt by hand for every tenant review.
Read the blueprint
What every Blueprint here guarantees
Zero prompt drift
Every Blueprint is a versioned contract. Run 1 and run 10,000 behave identically.
Scoped sessions
Each LLM call sees only the context it needs — no one giant prompt, no context rot.
Typed output
Blueprints return validated objects pinned to a schema, not text you have to parse.
Reusable like an API
Parameterise once and call it from anywhere — versioned, auditable, shareable.