Prompt operations screen with structured messages, variables, version history, diffs, evaluation scores, environment promotion, and rollback context.
Model playground
Side-by-side multimodel playground with reusable prompt messages, model and parameter controls, streaming responses, token and latency stats, and save-to-eval actions.

More like this
Browse Application UI
Inspectable response run with streaming, tool-use, waiting, completed, and failed states plus a concise reasoning summary that keeps private chain-of-thought hidden.

Tool-call and MCP control plane with server health, discoverable tools, typed request and response payloads, permission scopes, and human approval gates.

A human approval queue where each card names the automation, the action awaiting sign off, context chips, the requester, and approve or decline actions.

Vertical trigger, filter, and action step editor with dashed connectors, a Draft badge, and Save automation for a single linear flow.

Failed automation card with the error in plain words, a retry meter showing attempts used and remaining, and a payload replay affordance.

Template gallery with six categorised automation cards, use counts, a search field, and a start-from-scratch tile.

An automation analytics panel with four stat tiles for runs, success rate, time saved, and active count above a runs per day bar chart.

Branch decision panel with an If row of field, operator, and value chips, Then and Else path summaries, and an add condition affordance.

A connected apps panel where each row shows the account, a status badge, dependent automation count, and a reconnect action for broken connections.

Four-step dataset import workflow for selecting a project, uploading data, mapping fields, and confirming the import.

Automation run debugger showing the run header, a per step status list, and the input and output of the failed step expanded.
