Goal of this module
Take everything you built in test — entities, the complete-task and triage-task logic blocks, the crew, the automation — and ship it to production. Then read where usage and limits live.
Config moves, data doesn't
Promotion moves configuration — entities, logic blocks, events, settings, crews — from test to prod. App data never moves: prod starts with its own empty data spaces. This keeps prod changes intentional and reviewable.
1 · Provision production
Each org gets production on demand. Provisioning is idempotent — safe to call again.
POST /user/provision-prod2 · Diff test against prod
Before publishing, see exactly what will change.
GET /deploy/diff what differs between test and prod3 · Publish → activate
Publishing stages selected items; activating makes them live. If something's wrong, roll back in one step.
POST /deploy/publish stage selected items for promotion
POST /deploy/activate make staged items live
POST /deploy/rollback revert to the previous versionProduction is read-only in the console
When the console is switched to prod you can inspect what's live but not edit it. All changes are made in test and deployed — that's the guardrail that keeps prod intentional.
4 · Observability & limits
Usage is metered in real time. Your org-level usage — API calls, record counts against your plan — is available from the observability endpoint and surfaced in the console.
GET /observability your org's usage and plan stateLimits are enforced at the edge
Plan limits (API calls, record counts) are checked at handler entry. When an org crosses a threshold, flags like apiLimitExceeded / dailyLimitExceeded gate further writes — so usage isn't just reported, it's enforced.
Try it
Publish TaskFlow to prod, switch the console to prod and confirm the entities are there, then open the usage view and watch your call count from today's workshop.
You built TaskFlow
In one day you went from an empty org to a deployed, multi-tenant, AI-assisted, automated app — entities and CRUD, app-user auth, workspace isolation, ACL, a custom logic block, an AI crew, a Slack automation, and a production deploy. Every piece was configuration over a shared runtime — no servers, no backend deploys.