Start here
What is Agent Experience?
A practical definition of AX, where it sits beside UX and DX, and the questions it helps a team ask.
Read guideInsights
Practical AX guides for better products.
Pick the problem in front of your team—discovery, tools, documentation, workflows, evaluation, or control—and start with a useful, concrete guide.
Explore the field mapFramework
An Agent Experience (AX) framework
Five parts to look at, four steps to work through, and one rubric to inventory what you have declared — assembled into a framework you can run.
Read guideTools
Tool descriptions are product surfaces
The language around a tool helps determine whether an agent can find it, choose it, and use it well.
Read guideDocumentation
Make documentation legible to agents
Good agent-facing documentation makes intent, constraints, and recovery paths easier to use.
Read guideDiscovery
Build websites agents can understand
Clear structure, stable interactions, and meaningful labels help both people and agents navigate a site.
Read guideFoundations
AX, UX, DX, and GEO: where each fits
Keep the related disciplines connected without pretending they describe the same product problem.
Read guideAgent readiness
The Open Agent-Readiness Rubric
Thirty-six checks across six dimensions, weighted to 100, with the exact request or inspection that settles each one.
Read guideEvidence review
Does llms.txt actually work?
What the 2026 adoption and server-log studies actually show, and what they don't.
Read guideStandards
The state of agent-web standards
llms.txt, AGENTS.md, agents.txt, ARD, MCP server cards, WebMCP, A2A and Content Signals — compared on who reads them, not on who announced them.
Read guideDiscovery
Agentic Resource Discovery, implemented
What ARD is, why the well-known path changed in August 2026, and a copy-pasteable manifest for a content site.
Read guideTools
An agent’s tool catalog needs a pruning strategy
A smaller, clearer set of capabilities can be more useful than a long list of overlapping actions.
Read guideStrategy
Start with workflows, not autonomy
A bounded workflow is often the fastest way to learn where agentic behavior genuinely helps.
Read guideEvaluation
Evaluate decisions, not just answers
Inspect tool choice, handoffs, recovery, and the path to an outcome—not only the final response.
Read guideEvaluation
How do you measure agent experience?
Four families of AX measurement, the instrumentation each one needs, and the claims none of them can support.
Read guideEvaluation
Run an Agent Experience review
A repeatable way to inspect one meaningful workflow and choose the next product improvement.
Read guideDelegated action
Human approval is a workflow, not a pop-up
Consequential actions need a clear pause, a meaningful choice, and a usable route back into the work.
Read guideReliability
Design for recovery, not perfect runs
A helpful system knows when to stop, explain what happened, and make the next safe step clear.
Read guideInterfaces
A good agent dashboard shows the next decision
Operational surfaces earn their place when they explain what needs attention, why, and what to do next.
Read guideWorkflows
Every handoff should preserve the reason for the work
A handoff should carry the task, the decision, and the exact question that needs resolution.
Read guideReliability
Retries are product history, not a clean slate
Visible attempts reveal whether a system recovered cleanly or simply made people do more work.
Read guideEvaluation
Build an evaluation loop that improves the product
Use recurring workflow failures to guide durable product improvements instead of collecting a decorative score.
Read guide