Functions-as-a-service (Lambda, Cloud Functions, Azure Functions) run your code in response to events (HTTP request, queue message, upload, schedule) without you managing servers. You pay roughly for invocations + duration × memory; the platform handles scaling to zero and cold starts.
You upload a deployment package or point at a container image; the cloud injects an entry handler. Each invocation gets an isolated runtime (with caveats). Concurrency limits and timeouts cap runaway work — same instinct as Coffee Server’s Moo timeouts, but at hyperscale.
Serverless shines as glue: resize an image on S3 upload, drain a queue, authorize an API Gateway request. It is awkward for long-lived connections or huge in-memory workloads — then you pick containers or VMs.
COFFEE-FUNCTIONS (FUNCTIONS1-CORE.html) — “Edge Runtime” console mock: function list, editor chrome, logs. It teaches UX and vocabulary for the category.
Moo on Coffee Server is the crib-scale cousin: allowlisted plugins, worker execution, bounded time — same invoke → run → JSON shape without AWS billing.
Repo note: EXEC-PROXY-BODY-ISSUE.md documents a real integration sharp edge — good reminder that SaaS docs are full of those.