An independent field guide
A field guide to Agent Experience
Agent Experience (AX) is how well a product works when an AI agent uses it on someone’s behalf. These free guides cover the tools, docs, approvals and recovery paths that agents depend on, and how to get them right.
Written for product managers, designers, engineers, technical writers and AI teams putting an agent in front of a real workflow.
Start here
Write tool descriptions an agent can act on
Help an agent understand when a tool fits, when it does not, and how to recover from a bad input.
Tools & APIs
Keep context through the handoff
Carry the task, the decisions, and the next useful step forward when work passes between an agent and a person.
Workflows
Find the next thing to improve
Check one job your agent does against the Open Agent-Readiness Rubric. You get a profile across six tracks, each rated at one of four levels.
Evaluation
Look at the public work
The rules, findings and examples the guides draw on.
Checker rules
The rules of ax-check, a checker for tool descriptions, command-line tools and agent instructions, documented ahead of its release.
Evidence register
Public findings on how agents find, choose and use tools, each with its source and its evidence type (preprint, vendor report, public issue).
What changed in AX
A digest, about every two weeks, of the launches, specifications and research that matter to agent-facing products, each with its source.
Worked examples
Made-up but realistic agent workflows, each with its tests and its limits.
Use it in your assistant
Connect your assistant to this free, read-only MCP server and it can search and read the guides, fetch the rubric and run audit_agent_path, the one check you can run here today.
https://agentexperience.tech/api/mcpCommon questions
What is this site?
An independent guide to AX, written in public. Everything on it is free: practical guides, an open readiness rubric, worked examples, review templates and a read-only MCP server.
Who is it for?
Product managers, designers, engineers, technical writers and AI teams putting an agent in front of a workflow that matters.
Where should I start?
Read the guide that defines AX. Then take one workflow you want to hand to an agent and walk it through the field map. The rubric shows what to fix first.
What can I check for myself?
The rubric and its JSON copy, the tool-description test cases, the worked examples and their tests, and this site’s own agent-facing files.