Make your business discoverable to AI agents.

Can AI agents find your business?

Scan how AI agents can discover and understand the public machine interface of this hostname — APIs, tools, and machine-readable capabilities on the URL you enter.

Public URLs only · No signup · Passive GET/HEAD requests · Help: [email protected]

PUBLIC SIGNAL ANALYSIS● EVIDENCE-BASED
?AGENT VISIBILITY SCORE
DiscoverySCAN
MetadataVERIFY
ClarityEXPLAIN
TrustEVIDENCE
What the scan checks

Public signals, read as a system.

A missing optional standard is not a verdict. The scan combines deterministic validation, profile-aware scoring, and bounded semantic evaluation.

01

Discovery paths

Crawl policy, sitemaps, public docs, ARD draft linkage, and well-known resources.

02

Machine metadata

OpenAPI, A2A Agent Cards, structured MCP evidence, and public plugin packages.

03

Capability clarity

Whether public language clearly describes concrete actions, objects, examples, and limits.

04

Interoperability

Schemas, operation identity, authentication guidance, versioning, and invocation clarity.

05

Trust signals

HTTPS, identity consistency, security documentation, and exposed-secret detection.

06

AI understanding

A bounded semantic evaluator checks whether evidence supports the business and its capabilities.

How to read the score

90+ means agent-native, not merely well documented.

Agent Visibility Score uses Score Model v1.0, calibrated on 20+ live hosts with frozen weights and hostname-scoped evidence. A successfully crawled ordinary website typically lands near 40. One validated machine interface is 70–75. Multiple validated protocols sit in the mid-80s. 90–100 is Agent-native, including ARD 0.9 draft. The commercial path is 40 → 70 → 85 → 90. Blocked crawls are labeled Insufficient evidence and are excluded from ranking calculations.

90–100 · Agent-native

Comprehensive structured discovery, including emerging standards such as ARD 0.9 draft.

75–89 · Strong

Clear public agent/API evidence, with remaining gaps in discovery, metadata, or trust.

40–74 · Human-readable to one spec

Ordinary sites land near 40. One validated machine interface is 70–75. Marketing hosts without a contract sit near 30.

Insufficient evidence

Blocked or empty crawls. The raw score is stored and excluded from percentile and ranking calculations.

How it works

Scan, fix one contract, then keep the verified score honest.

MachineReach is not an uptime monitor. Use it to see the public machine interface, repair one artifact, rescan, share the verified score, and watch that verified state for regressions.

01

Scan

Get a public report and a basic score explanation.

02

Fix

Follow the exact remediation plan for the next score band.

03

Verify

Rescan. The new snapshot is the verified score.

04

Share

Use the public verification page and badge.

05

Monitor

Get email only if verification degrades, cannot be refreshed, or recovers.

Pricing

Scan freely. Pay when you need the loop to keep running.

Billing is not self-serve yet. The product boundary is already the one customers feel: a one-off scan versus protecting a verified hostname.

Free

  • Scan a public hostname
  • Public report
  • Basic score explanation

Paid

  • Exact remediation plan
  • Hostname history
  • Verified badge
  • Continuous monitoring
  • Regression and recovery alerts

Want paid access for a hostname? Email [email protected].