Agents AI UX patterns
Agent patterns cover tool use, autonomy budgets, approvals, checkpoints, and orchestration—where users need to see what the system did and revoke or roll back.
Related framework: Agentic UX framework
Start here
Core patterns for agents UX.
30 patterns
Plan & Execute
Breaking goals into steps
Tool Use
Visualizing AI using external tools
Workflow Builder
Visual drag-and-drop workflows
Memory Management
Viewing what AI remembers
Task Queue
Visual queue of agent tasks
Human Handoff
Escalating to humans
Escalation Thresholds
Auto-demote autonomy when risk crosses a line
Error Recovery Strategies
Configurable retry/fallback patterns
Self-Correction
Agents fixing errors
Prompt Chaining
Multi-step logic
Conditional Logic & Branching
Branch workflows by conditions
Scheduled Tasks & Recurring Actions
Time-based automation triggers
Multi-step Forms with AI
Adaptive progressive forms
Approval Workflows
Human approval gates
Agent Orchestration
Visual flow for multiple agents
Agent Performance Metrics
Dashboard showing agent success rates
Agent Versioning
A/B test different agent configurations
Agent Marketplace
Browse and install pre-built agents
Blast Radius Visualization
Preview scope before agent executes
Reversibility Marking
Label actions by how easily undone
Time-Delayed Execution
Cancelable countdown before high-impact actions
Suggest / Confirm / Execute
Three autonomy modes for AI agents
Pre-Task Cost Estimate
Forecast tokens, dollars, and time before run
Checkpoints and Restore
Named snapshots with one-click restore
Per-Action Autonomy
Autonomy scoped per capability, not per app
Autonomous Mode Display
Persistent signal when agent runs unattended
Autonomy Budgets
Time- or action-bounded unattended runs
Context Portability
Legible payloads across app boundaries
Ambient Presence Displays
Low-attention signals for agent state
Sandbox Preview
Dry-run plan with explicit side effects before execute
Frequently asked questions
How are agent patterns different from chatbot patterns?
Chat patterns focus on conversation contract and turns. Agent patterns focus on side effects, tool calls, permissions, checkpoints, and durable control across sessions.
What should users always see before an agent acts?
Intent, scope, and reversibility: which tools will run, what data they touch, and how to stop or undo. Approval workflows and autonomy budgets make that contract explicit.
When do I need a task queue versus a single plan-execute flow?
Use plan-execute for linear multi-step work. Add task queues when several operations run in parallel or over time and users need status per task, not only a final message.
How do I design rollback and checkpoints?
Expose checkpoints before destructive steps, label what state will be restored, and keep rollback one gesture away. Users forgive agent mistakes when recovery is obvious.
Which products exemplify agent UX in this catalog?
Pattern examples reference Cursor, Devin-style agents, workflow builders, and assistants with tool pickers—open any card’s examples list for specific references.
Where is the Agentic UX framework?
Use the linked framework for territories (permission, rollback, autonomy) and anti-patterns. Category patterns here are the implementation catalog mapped to those territories.