# The Six Foundational AUX Patterns

> Canonical: https://auxfirst.com/#patterns
> License: CC BY 4.0 · auxfirst agency 2026

Repeatable interaction patterns that help teams deliver agentic behaviour with precision and consistency. These are architectural choices you can layer, remix, and extend. Almost every agentic feature you'll ever design is some combination of these six.

> Principles are what you believe. Heuristics are how you evaluate. **Patterns are what you actually ship.**

## Pattern 01 · Intent Handshake

A short exchange that clarifies the user's goal before the system acts. The agent restates what it thinks you want, names its assumptions, and offers a chance to redirect before any cost is incurred. Avoids one-shot prompting where agents jump to conclusions and erode trust. Establishes mutual alignment.

**Use when:** the action is non-trivial, expensive, irreversible, or sensitive.
**Anti-pattern:** auto-executing on a half-understood request.

## Pattern 02 · Confidence Cues

Make reasoning visible. Expose sources, surface uncertainty, show the logic — but *tapered*, not overwhelming. Help users calibrate trust. Agents that show their work feel self-aware. Agents that don't feel mysterious — and mysterious is one bad output away from untrustworthy.

**Use when:** the agent produces an output the user might act on.
**Anti-pattern:** showing a raw confidence percentage. Users don't have the statistical intuition to act on it.

## Pattern 03 · Adaptive Canvas

The interface reshapes itself in response to the evolving task. The right tools surface when needed. Previews appear when relevant. A stable substrate that adapts at the edges while preserving spatial memory. The product feels like it thinks alongside the user.

**Use when:** the task surface changes shape mid-flow.
**Anti-pattern:** a static UI bolted onto an agentic backend.

## Pattern 04 · Escape Hatch

Every agentic system needs obvious ways to undo, revise, or override. The Escape Hatch is what gives users psychological safety to engage in the first place. When people know they can recover, they explore more, delegate more, and trust more.

**Properties:** always one click, always visible, always unambiguous about what it cancels.

**Use when:** the agent has authority to act.
**Anti-pattern:** a single global "stop" button that doesn't tell the user what it stops.

## Pattern 05 · Memory in Motion

Recall of contextual relevance across time — past decisions, preferred formats, recurring tasks, corrections. Must be **transparent and editable**. Building memory governance is where the real design work lives. Shifts the relationship from convenience to continuity.

**Use when:** the agent operates across sessions.
**Anti-pattern:** exposing "memory" as a chat transcript. That's logs, not memory. Real memory editors surface the *abstractions* the agent has formed about the user.

## Pattern 06 · Generative Momentum

The agent initiates — drafts, outlines, candidate queries — and invites the user to shape them. The blank page disappears. Co-authorship replaces command-and-execute. Must be paired with Pattern 01 and 04 to stay safe.

**Use when:** the user benefits from a starting point.
**Anti-pattern:** generating finished output as the default, without scaffolding for revision.

## The contract underneath every pattern

Patterns are not components. A component is reusable code; a pattern is a reusable **expectation**. Each pattern describes:

- **The shape** — what the user sees
- **The contract** — what the agent and the human each owe
- **The failure mode** — the most common bad version
- **The trigger** — when to deploy it

Skip the contract and the pattern degrades into decoration.

## The expanded catalogue (24 patterns)

The six above are the highest-frequency primitives. The complete pattern catalogue extends to 24 patterns across four families:

### Trust family
Confidence ribbon · Source citation · Reasoning trace · Calibration cue · Dissent surface · Version diff

### Control family
Permission scope · Autonomy spectrum · Escape hatch · Kill switch · Dry-run mode · Budget governor

### Memory family
Memory editor · Context inspector · Identity card · Forget control · Scope boundary · Recall trace

### Orchestration family
Activity stream · Approval queue · Intervention point · Audit trail · Hand-off · Parallel session view

The full catalogue is at https://auxfirst.com/agent-first-design-patterns.html

## Mapping patterns to autonomy

Patterns are not optional based on taste. They are **required** based on where the agent sits on the autonomy spectrum. Each step up the spectrum demands additional patterns:

| Autonomy mode | Required trust patterns | Required control patterns | Required orchestration |
|---|---|---|---|
| Human-only | Confidence ribbon, reasoning trace | Permission scope | Activity stream |
| Approve-each-action | + source citation, version diff | + escape hatch | + approval queue |
| Review-before-act | + dissent surface | + dry-run mode | + intervention point |
| Act-and-notify | + calibration cue | + budget governor | + audit trail |
| Autonomous | All of the above | + kill switch | + hand-off, parallel view |

Read the table as a build order. A product moving an agent up the spectrum doesn't need new model capabilities — it needs new patterns.

---

Related: [The AUX Manifesto](https://auxfirst.com/manifesto.md) · [The ten AUX heuristics](https://auxfirst.com/heuristics.md) · [The complete pattern catalogue](https://auxfirst.com/agent-first-design-patterns.html)