ToolJet has published another worked build in its MCP series, and this one is aimed at the people least likely to trust it. In Build a Live Incident Room App with ToolJet MCP and Jev, a coding agent connected through ToolJet MCP generates a two-page app for a payments on-call team: a war room with open incidents, the selected record, its first three alerts, a triage card, commander controls and a live timeline, and a review page for agreement, overrides, time to acknowledge and time to resolve. Each open incident is triaged by Jev, TypeSafe’s decision model, through a datasource, returning severity, blast radius, customer impact, page-leadership probability and likely cause “while a person decides.”
There are no performance figures, and the honest detail is the bug. “The customer-impact card showed the wrong scale, so the number looked lower than the triage meant. Jev returned the score as a 0-based legend index, not a 0 to 1 value.” That is exactly the class of error a generated app ships with, a field whose meaning was assumed rather than read, and the post fixes it in the open. The real build “took several passes,” consolidated afterwards into one reproducible prompt. The governance is specific: three workspace groups, commanders who can confirm or override severity, engineers who can only add timeline updates, and viewers with every action disabled, with “these write queries restricted to them on the server.” The design keeps the model’s opinion and the commander’s decision as separate records, so the review page can measure how often they agree.
The catalog reads the platform the agent builds on. The ToolJet provider page lists 4 API pages, and the post’s role model maps to the ToolJet Groups API and the ToolJet Users API, with the generated app itself under the ToolJet Applications API and the ToolJet Workspaces API. The agentic access profile maps 23 operations, 15 of them acting. Five Agent Readiness dimensions are lit: spec presence, agentic access, the MCP server, auth clarity and the rate limit signal.
The Kin Score is 44.4, developing band, with discoverability at 65.0 and contract quality at 56.4 against developer ergonomics at 23.8 and contract governance at 23.5. The Agent Readiness score is 21.0, agent-aware. The catalog records authentication as a static access token passed in a Basic header, and delegated identity, protected resource metadata and dynamic client registration are all unlit. The incident room separates a commander from an engineer from a viewer carefully, inside the app. The API that the agent used to build it does not yet describe a way to tell one caller from another. ToolJet has written role-based access into what the agent generates. The contract the agent generated it through is still a single key.