Agent readiness test lab · updated 2026-08-30
10 free public tools turn a vague “agent-ready” claim into a bounded diagnostic or plan, inspectable evidence, and a result you can compare after every release.
One shared execution model
Six focused entry points
Scan a public URL, inspect a local JSON artifact, or build the task contract first. Each route states what it observes and what it cannot prove.
Crawler and browser access preflight
Compare robots policy, simulated agent user-agent responses, server-rendered content, browser stability, CAPTCHA, and security evidence in one public-site preflight.
Live public scan projection
Open Access MatrixLocal Lighthouse evidence importer
Import a Lighthouse JSON report locally, preserve official audit IDs and display values, and turn observed failures into an agent-journey rerun checklist.
Local file analysis
Open Lighthouse ImporterGoal contract builder
Define the goal, starting state, allowed and prohibited actions, safe stop, assertions, and authoritative readback before an agent touches a site.
Planning and contract builder
Open Journey ContractCommerce readiness preflight
Check the public substrate an agent needs before checkout: product data, structured prices, price parity, policies, HTTPS, and discoverable interaction evidence.
Live public scan projection
Open Stripe Link PreflightLocal tool-registration linter
Lint proposed WebMCP-style tool registrations for names, descriptions, input schemas, constraints, consequences, harmless-call clarity, and declared human-form parity.
Local file analysis
Open WebMCP PreflightComparable replay plan
Create one content-versioned task contract and four explicit result slots for Grok, Claude, ChatGPT Work, and Gemini so manual or future authorized runs remain comparable.
Planning and contract builder
Open Cross-Agent ReplayPublished now
Analyze robots.txt for named crawler tokens, then inspect bounded public-page delivery, redirects, X-Robots-Tag, and HTML meta robots.
/api/tools/crawler-accessReview OpenAPI JSON for operation identity, descriptions, auth and scopes, errors, examples, retry and rate-limit guidance, lifecycle, pagination, and workflows.
/api/tools/openapi-readinessValidate an A2A Agent Card shape and surface missing identity, version, skill, supported-interface, media-mode, and capability details.
/api/tools/a2a-agent-cardSupporting utility
This existing helper stays available as a browser workflow. It is intentionally omitted from AgentReady's OpenAPI, MCP, and machine-tool inventories until its endpoint shares the same versioned contract and abuse controls as the published diagnostics.
Inspect the file served at the origin root, catch HTML fallbacks, and generate a maintainable starting document.
Gated beta work
These capabilities handle sessions or live protocol behavior, so they are not anonymous toys. The published pages describe the intended boundary without pretending the production runner is available.
Define the bounded, read-only contract for a future reference-agent journey with an expiring session handoff and deterministic end-state assertions.
Exercise safe discovery and read-only MCP behavior separately from security and conformance claims.
From visibility to actionability
A crawler can fetch a page while a browser agent still fails to understand controls, satisfy authentication, choose the right variant, or verify that a write action succeeded. The suite keeps those claims separate so a passing robots check never masquerades as end-to-end task completion.
See the opt-in, category-specific comparison model and the evidence rules required before a public ranking.
Explore cohortsBrowse timestamped public-company journey profiles with exact tasks, observable evidence, and safe stopping points.
Open the indexSee the post-login journey foundation, policy boundaries, evidence model, and gated execution roadmap.
Review the platformNo. The public tools do not sign in, submit consequential forms, purchase anything, or mutate your site. They produce bounded diagnostics and planning artifacts from public scans or files you analyze locally in the browser.
No. It is a sandbox-readiness preflight. It inspects observable commerce signals and produces a test contract without entering payment credentials, confirming a payment, or creating an order.
No. It exports one comparable replay plan with every provider row marked not_run. Running provider accounts, browsers, or harnesses remains a separately authorized step.
Every tool uses the same execution vocabulary: a GoalContract defines the task and boundaries, a TestRun records what was attempted, Evidence supports the result, an OutcomeReceipt states what was observed, and a Diff compares compatible receipts.
Define the journey, its permission boundary, the safe stopping point, and the readback that would count as success before asking any agent to act.
Build a GoalContract