Agent Experience

An independent field guide to Agent Experience

Agent Experience (AX): research, design, and engineering.

Agent Experience (AX) is the quality of a product as AI agents experience it while acting for a person. Practical guidance for the tools, documentation, workflows, approvals, recovery paths, and evaluations those agents rely on.

Start with the guide

Explore all guides, examples, and work →

Written and curated by Himadri Mishra.

The starting point

Good models still need well-designed products.

AX looks at how an agent discovers a capability, understands the instructions, gets authority, executes safely, reports evidence, recovers from uncertainty, and hands work back to a person.

Explore the AX framework

A few places to begin

All guides
Tools & APIsWrite tool descriptions an agent can act onHelp an agent understand when a tool fits, when it does not, and how to recover from a bad input.WorkflowsKeep context through the handoffCarry the task, the decisions, and the next useful step forward.EvaluationFind the next thing to improveUse a public rubric to check one job your agents do.

Inspect the public work

Selected public workSee inspectable artefacts behind this field guide, including the site MCP, rubric, fixtures, and public source library.Worked examplesStudy synthetic AX workflows with state flows, deterministic tests, and explicit limits.What changed in AXA roughly fortnightly digest of agent-facing launches, specifications and preprints, each with its source.AX evidence registerPublic findings behind the guidance, with sources, evidence classes and limitations.

Keep exploring.

Find reusable resources, public guides, and agent-facing artefacts without starting from a sales page.

Browse the resources

Use the guides in your assistant.

Connect the public MCP server to search the guides, retrieve the rubric, and run a bounded catalogue diagnostic.

Connect to the MCP

Common questions

Read the full definition
What is this site?

Agent Experience is an independent field guide to AX, written and curated in public. It publishes practical guides, an open agent-readiness rubric, synthetic worked examples, copyable review templates, and a read-only MCP server, all free to read and inspect.

Who is the field guide for?

Product managers, designers, engineers, technical writers, and AI teams who are putting an AI agent in front of a workflow that matters and need the product surfaces around it to hold up.

Where should I start reading?

Begin with the definition guide, then take one workflow you already want to delegate and walk it through the field map. The readiness rubric then shows which parts of that job are ready, and what to fix first.

What can I inspect here rather than take on trust?

The Open Agent-Readiness Rubric and its JSON twin, the MCP tool-description fixture, the worked examples with their deterministic tests, and this site’s own agent-facing files.