← LibraryCollection
Agentic flows
How the model shows its work, in the world.
When the model plans, uses tools, files, browsers, APIs, the interface has to show a whole second loop. Most products hide it. The best ones stage it.
12 patterns
- 12 minThe plan previewHow to show the model's intent before it spends your time.
- 10 minThe tool call, visibleWhen and how to show the model reaching for a tool.
- 9 minThe interrupt handleHow the user stops a running agent without losing state.
- 10 minThe diff previewBefore-and-after for anything the agent is about to change.
- 8 minThe approval checkpointA pause-for-human-signoff, named and scoped.
- 8 minThe step receiptAn after-the-fact trail of what the agent actually did.
- 9 minThe rollback affordanceUndoing what the agent did, when undo is possible.
- 9 minThe rate-limit displayMaking invisible quotas visible without anxiety.
- 10 minThe model fallbackChoosing a cheaper model without surprising the user.
- 11 minThe macro recorderTurning a good agent run into a reusable recipe.
- 10 minThe cost tickerA running total the agent can't hide from.
- 11 minBackground task dockWhere long-running work lives while you keep typing.
Other collections
- The prompt surfaceHow the box asks for intent.→
- The responseHow the model performs the answer.→
- Memory & contextWhat the model carries with it.→
- Trust & evidenceHow the model earns the reader's belief.→
- Multimodal & artifactsWhen chat isn't enough.→
- CollaborationWhen the session has more than one human in it.→
- OrchestrationHow many agents, at what cadence, inside what ceiling.→
- Dev & evalThe surfaces that make AI products debuggable.→