# Agent Experience (AX) > Agent Experience (AX) is the practical work of making products, documentation, tools, and workflows easier for AI agents to discover, understand, use, recover through, and act within on behalf of people. Agent Experience is an independent field guide for product, engineering, design, documentation, and AI teams building agent-facing products. ## Start here - [What is Agent Experience?](https://agentexperience.tech/insights/what-is-agent-experience/): A practical definition of AX and the questions it helps teams ask. - [An Agent Experience (AX) framework](https://agentexperience.tech/insights/agent-experience-framework/): The AX framework used across this site, stated in one place: a five-part field map (orient, discover, act, collaborate, learn), a four-step method for applying it to one workflow, and the Open Agent-Readiness Rubric as the inventory layer. Includes an explicit disambiguation from Ax, the unrelated open-source DSPy-style LLM framework for TypeScript. - [Explore every page](https://agentexperience.tech/explore/): The complete directory of guides, examples, resources, work and contact paths. - [The Agent Experience field map](https://agentexperience.tech/research/): A five-part operating map: orient, discover, act, collaborate, and learn. - [All Agent Experience guides](https://agentexperience.tech/insights/): The complete practical guide library. - [A practical approach to AX](https://agentexperience.tech/method/): A four-step way to improve one real workflow. ## Discovery and documentation - [Build websites agents can understand](https://agentexperience.tech/insights/agent-friendly-websites/) - [Make documentation legible to agents](https://agentexperience.tech/insights/make-documentation-legible/) - [AX, UX, DX, and GEO: where each fits](https://agentexperience.tech/insights/ax-vs-ux-dx-geo/): A side-by-side comparison of AX, UX, DX, and GEO, plus disambiguation of the two AI-sense definitions of Agent Experience and the older contact-centre meaning. - [Does llms.txt actually work?](https://agentexperience.tech/insights/state-of-llms-txt/): An evidence review of llms.txt as of September 2026: adoption estimates disagree by a factor of five, AI search crawlers rarely fetch the file, Google Search says it ignores it, and coding agents are the one demonstrable consumer. - [The state of agent-web standards](https://agentexperience.tech/insights/agent-web-standards/): A September 2026 comparison of ten agent-web standards. Of these, AGENTS.md, MCP and A2A have real consumers; Content Signals has real deployment; llms.txt has publishers but very few readers; ARD, WebMCP, MCP server cards and agents.txt are early-stage or contested. Each entry lists steward, spec URL, status, evidence of consumption, and a recommendation. - [Agentic Resource Discovery, implemented](https://agentexperience.tech/insights/agentic-resource-discovery/): Agentic Resource Discovery (ARD) is a proposal-stage open specification, authored by contributors from Google, Microsoft and Hugging Face and announced 17 June 2026, for publishing a machine-readable inventory of agents, tools and other resources on your own domain. As of v0.91 (26 August 2026) the manifest lives at /.well-known/ard.json; /.well-known/ai-catalog.json is the predecessor path. ## Tools and interfaces - [Tool descriptions are product surfaces](https://agentexperience.tech/insights/tool-descriptions/) - [An agent’s tool catalog needs a pruning strategy](https://agentexperience.tech/insights/agent-tool-catalogs/) - [A good agent dashboard shows the next decision](https://agentexperience.tech/insights/show-the-next-decision/) ## Workflows, trust, and recovery - [Start with workflows, not autonomy](https://agentexperience.tech/insights/start-with-workflows/) - [Every handoff should preserve the reason for the work](https://agentexperience.tech/insights/preserve-context-through-handoffs/) - [Human approval is a workflow, not a pop-up](https://agentexperience.tech/insights/approval-is-a-workflow/) - [Design for recovery, not perfect runs](https://agentexperience.tech/insights/design-for-recovery/) - [Retries are product history, not a clean slate](https://agentexperience.tech/insights/retries-are-product-history/) ## Evaluation and improvement - [The Open Agent-Readiness Rubric](https://agentexperience.tech/insights/agent-readiness-rubric/): A vendor-neutral v0.1 rubric of 36 criteria across six dimensions — discovery, structure and semantics, machine-readable content, action safety, recovery, and policy signals — each with a concrete check, a rationale, and a weight totalling 100. - [Evaluate decisions, not just answers](https://agentexperience.tech/insights/evaluating-agent-workflows/) - [How do you measure agent experience?](https://agentexperience.tech/insights/measure-agent-experience/): A guide to measuring Agent Experience as a measurement-design problem: name the construct, pick from four families (declared surfaces, path quality, reliability under repeats, and cost of the path), instrument the run record first, and keep the claim inside what the construct supports. Describes public measurement objects — readiness scans, transcript-derived scores, trajectory metrics, pass^k — without ranking them. - [Run an Agent Experience review](https://agentexperience.tech/insights/run-an-ax-review/) - [Build an evaluation loop that improves the product](https://agentexperience.tech/insights/build-a-reliable-agent-evaluation-loop/) ## Practice and evidence - [Selected work](https://agentexperience.tech/work/): Publicly inspectable work with contribution and limits; not a collection of client outcome claims. - [Worked examples](https://agentexperience.tech/examples/): Synthetic capability preflight and recovery examples, deterministic tests, exercises and Markdown twins. Not benchmarks. - [Lifecycle crosswalk, Markdown](https://agentexperience.tech/research/index.md): Seven workflow stages mapped to the five product systems. An editorial method, not an industry standard. - [Capability and authority preflight before action, Markdown](https://agentexperience.tech/examples/capability-authority-preflight/index.md): A synthetic billing request shows how an agent should split the job, inspect exposed tools, verify authorization, and stop before a destructive call it is not allowed to make. - [Timeout recovery: verify before retry, Markdown](https://agentexperience.tech/examples/timeout-recovery-verify-before-retry/index.md): A synthetic sync job shows how an agent should treat an uncertain timeout as product history, verify state, and avoid repeating a side effect blindly. ## Reference - [Agent Experience (AX) glossary](https://agentexperience.tech/glossary/): Forty-three Agent Experience terms defined, with origins and dates where a term has one. - [Agent Experience resources](https://agentexperience.tech/resources/): Ungated templates and references for reviewing one agent workflow, including the AX Workflow Review brief and rubric links. - [AX Workflow Review brief, Markdown](https://agentexperience.tech/resources/workflow-review-brief.md): A copyable one-workflow template for mapping discovery, retrieval, tool execution, approval, recovery, handoff and the next product test. - [MCP tool-description test cases](https://agentexperience.tech/resources/mcp-tool-description-test-cases/): A copyable fixture for checking target intent, neighbouring tools, schema inputs, result shape and recovery across this site's four compatible reference MCP tools. - [MCP tool-description test cases, Markdown](https://agentexperience.tech/resources/mcp-tool-description-test-cases.md): Downloadable version of the MCP tool-description fixture. - [The Open Agent-Readiness Rubric, machine-readable](https://agentexperience.tech/agent-readiness-rubric.json): The rubric's 36 criteria, checks and weights as JSON. ## Work with Agent Experience - [AX Workflow Review](https://agentexperience.tech/services/): Diagnosis, design and engineering, or evaluation and enablement for a bounded workflow. Scope is agreed before work; no guaranteed rankings, citations or agent success. - [About](https://agentexperience.tech/about/): Public author profile and project boundaries. - [Contact](https://agentexperience.tech/contact/) ## Optional - [Full text of every guide](https://agentexperience.tech/llms-full.txt): One markdown file containing all 21 guides. - [RSS feed](https://agentexperience.tech/rss.xml): New and updated guides. - [MCP server](https://agentexperience.tech/api/mcp): Model Context Protocol server with audit_agent_path for bounded static public-page and sanitized tool-contract checks, plus the public reference tools. No private repo access or action execution. Streamable HTTP, stateless, POST only, no authentication. - [Connect and review a workflow](https://agentexperience.tech/connect/): Connect once and request an audit_agent_path diagnostic: evidence, proposed fixes and regression instructions. No agent-success score or runtime replay. The optional workflow_review prompt remains available for assistant-led discussion. - [ARD manifest](https://agentexperience.tech/.well-known/ard.json): This domain's invocable and machine-readable resources, per the Agentic Resource Discovery spec. - Every guide also has a markdown twin at `index.md`.