Everything auxfirst publishes openly — the manifesto, the principles, the heuristics, the pattern library, the open-source TrustKit, the dispatches, and the knowledge base. Definitions are CC BY 4.0. Executable tooling is MIT. No vendor wall, no "request a demo." Take what is useful.
The arguments under everything else. Why agentic UX is a discipline, not a feature. The eight beliefs we test every design decision against. The ten heuristics for evaluating agents the way Nielsen evaluated forms. Read these once a quarter.
The one-page argument — why software with memory, initiative, and judgment is not a tool but a collaborator, and why that rewires the whole UX discipline.
Read on the home page → 8 PrinciplesTapered transparency, two-way prompting, clarify before commit, pushback as professional, multimodal for the moment, loop in experts, context as memory, personalization as moat.
Read the principles → 10 HeuristicsThe AUX answer to Nielsen Norman's classic usability heuristics — reimagined for agents. Optimized for quality of relationship, not clarity of screen.
Read the heuristics →The reference library — every concept, pattern, and platform with its own page. The knowledge base is the dictionary. The pattern library is the grammar. The agent-native platforms list is the inventory. Use these when you are auditing, briefing, or arguing.
Agent Experience, Agentic User Experience, AI UX and agent-first design are four different design problems with four different owners. The comparison table, a worked example, and a decision tree that tells you which one you actually have.
Compare the four terms → ReferenceEvery AUX concept, pillar, and supporting term — defined, contextualized, and cross-linked. The auxfirst dictionary, organized as a navigable cluster.
Open the knowledge base → PatternsThe six foundational interaction patterns — Intent Handshake, Confidence Cues, Adaptive Canvas, Escape Hatch, Memory in Motion, Generative Momentum — with the contracts they imply.
Browse the patterns → Pillar Guide · Section 01The master pillar guide — the discipline of building products where AI agents are primary actors and humans supervise. Process, patterns, autonomy spectrum, checklist, mistakes.
Read the pillar → Pillar Guide · Section 03The complete catalog of 24 patterns across four families — trust, control, memory, orchestration. The agentic equivalent of the form, the table, and the modal.
Open the catalog → InventoryThe working register of platforms designed natively for agents — not retrofits. Sortable, comparable, updated as the category settles.
Browse the register → Reference Index16 hosted agent platforms across AI, cloud, data, and developer infrastructure — benchmarked against Claude Managed Agents. Direct links to the right docs, not just homepages.
Open the index → GlossaryThe shared vocabulary of the agentic era — terms, clusters, and the language buyers and builders need to actually mean the same thing on Monday.
Open the glossary →Design defines the scope of delegation. Architecture describes the system that performs the work. Platform describes the capabilities it operates. Lifecycle describes how it is built, run and changed. Read them in order, or start where your question is.
What work to delegate to an agent, and how people supervise, steer, approve and override it. Principles, the seven-step process, interface patterns and the autonomy map.
Read the guide → 02 · Architecture · NewA practical reference model for the system that performs delegated work: eight areas of responsibility, the authority and approval flow, failure handling, six patterns and a worked revenue operations example.
Read the reference model → 03 · Platform · NewWhat makes a platform operable by agents: eight capabilities under four questions, the ten-question Agent-First Platform Test as a browser worksheet, and build, buy or combine criteria.
Run the platform test → 04 · LifecycleHow the system is built, evaluated, released, run and retired — the agentic successor to SDLC, CI/CD and SLOs, with a 24-question readiness check.
Read the field guide →The newsletter — long-form arguments from the agentic frontier. Where AGT sits relative to AUX. The agent-first API. The family office case. Each piece earns its place by changing how a reader thinks about an agent the next time they see one.
The full archive — agentic UX, agent-first design, AX, ad tech, family offices, and the comparisons that name the seams between disciplines.
Open the archive → SubstackThe same essays delivered as they ship. Roughly monthly. No marketing emails, no productized funnels — just the next argument.
Subscribe ↗Frameworks without schemas are think-pieces. The canon ships as runnable, citable, forkable artifacts — TrustKit on GitHub, the Trust Canvas on Figma. Definitions in YAML. Tooling under MIT. No vendor wall.
Heuristics, trust architecture, trust-gap taxonomy, agent specs, memory policies — all as YAML schemas. Forkable, citable, no vendor lock-in.
Open the repo ↗ auxfirst.com · Free toolOpen an AGENTS.md, llms.txt, brand.md or any Markdown file in the browser. Rendered and source views, a generated outline, and export to PDF, HTML or plain text. Client-side only — nothing is uploaded.
Twelve structural checks on an AGENTS.md, llms.txt or brand.md — missing boundaries, vague directives, instruction-shaped content, third-party links. Client-side only; nothing is uploaded.
Twenty-one questions, eight minutes: how far an agency actually trusts its agents, scored across authority, oversight, traceability, failure design and ownership. Runs in the browser — no email, no sign-up, only the anonymous score is counted.
Take the benchmark → Figma Community · FreeA one-page artifact for mapping an agent's trust topology — autonomy boundaries, handoffs, patterns per moment. Lean Canvas, for AUX.
Open in Figma ↗ agentoperability.com · Free ebookThe agent operability field guide: the three-layer model, the Action Heat Ladder, the Agent Identity Card, the 90-day path, and the scored 42-item self-assessment. Read online or download — no email wall.
Read the field guide ↗ Field guide · NewWhy enterprise agent pilots stall between demo and adoption — Geoffrey Moore's chasm mapped onto your org chart, and the five-layer whole product that carries adoption across. Free, 11 chapters, no email wall.
Read the guide → Field guide · NewFrom prompt to accountable system. The agentic successor to SDLC, CI/CD, GitOps, CMDB, SBOM and SLOs — nine parts, ten maps of an agent estate, an artifact index, and a 24-question readiness check that runs in your browser. Free, no email wall.
Read the guide →The same frameworks, run against one industry end to end — which agents to build, how hot each action is, and the order to build them in. Written to be forwarded to the people who would own the result.
Thirty agents across eight revenue lanes, sorted by two questions — does it compound, and how hot is its action. The format differences from web to broadcast, the join problem, an eighteen-month roadmap and a ten-question self-score.
auxfirst is the agency for the agentic era. We design AI systems that remember, adapt, and earn trust — Blueprint Sprints, Agent Experience Audits, Advisory Retainers. If the canon answered a question, the engagements answer the next one.