---
title: "Harness agent readiness score: 51/100 · Agentrail"
description: "The website is strongest in agent discovery, with its brand name and developer resources discoverable, but it lacks an MCP server or manifest and a published OpenAPI specification."
canonical: "https://www.agentrail.co/agent-readiness/harness.io?h=0N04C40JEANGE00W"
---

Full report

## See what each check found, and how to fix it.

Evidence, step-by-step fixes and your score over time, for your own site.

[Get started →](https://app.agentrail.co/login?utm%5Fsource=agentrail.co&utm%5Fmedium=website&utm%5Fcampaign=agent-readiness%2Fharness.io&utm%5Fcontent=company-report)

Free to start. No card.

Usability

21 to fix

- OAuth 2.0 supportRequiredHow to fix →
- Scoped permissionsRecommendedHow to fix →
- Agent onboarding frictionRecommendedHow to fix →
- Web Bot Auth directoryEmergingHow to fix →

Fixes unlock with the full report

Where the score comes from

46 of 90 points

How harness.io scores on each of the four questions agents ask.

Discovery18/20

Access21/30

Usability7/40

Paymentsnot scored yet

Strongest

### Discovery

ARD discovery and Wikipedia / Wikidata entity presence are missing.

18/20

Most to gain

### Usability

OAuth 2.0 support, Scoped permissions, and 11 more are missing.

7/40

Most to gain

### Access

Content without JavaScript, Agent instruction / when-to-use, and 2 more are missing.

21/30

Not scored yet

### Payments ( bonus )

Counts when scoring opens.

—

Agent path

## How an agent moved through the site

We asked an agent a question about harness.ioand recorded every request it made.

**H**harness.ioAnsweredRan Nov 20, 2026

Question

I want to start using harness.io (harness.io). Find its setup and getting-started guide.

8

Steps

1m 46s

Duration

8

Pages opened

1

Failed

1 of 3 addresses the agent typed itself didn’t load. Every link it followed loaded.

1. 1harness.io /Named in the question _→_ **Read**
2. 2/supportA link on a page it read _→_ **Read**
3. 3harness.io /Named in the question _→_ **Read**
4. 4harness.io /Typed by the agent _→_ **Didn’t load · 301**
5. 5/docsA link on a page it read _→_ **Read**
6. 6/docs/getting-startedTyped by the agent _→_ **Read**
7. 7/docs/platform/get-startedTyped by the agent _→_ **Read**
8. 8/harness-platform/new-to-harness-platform/get-started.mdA link on a page it read _→_ **Read**

8/8 steps

StepsHow it found itCame back as

Step 8 of 8

`developer.harness.io/harness-platform/new-to-harness-platform/get-started.md`

How it got here

A link on a page it read

Response

200

Came back as

Markdown

88%

Pages that loaded

38%

Followed site links

100%

Read as markdown

How it found each page

Named in the question**2**

A link on a page it read**3**

Typed by the agent**3**

A search result**0**

What it reached

Home page**Step 1**

Pricing**Never reached**

Docs**Step 5**

Its answer

Answers aren’t shown publicly

See what agents answer about your own site.

Run agents on your site and read every answer, with every page they opened.

[Get started →](https://app.agentrail.co/login?utm%5Fsource=agentrail.co&utm%5Fmedium=website&utm%5Fcampaign=agent-readiness%2Fharness.io&utm%5Fcontent=journey-answer)

Free to start. No card.

Every check

139checks

- Passed**23**
- To fix**19**
- Not in the score**97**

42 checks count toward the score. The other 97 are tracked for later.

Discovery18 of 201 to fix1 partly5 passed

ARD discoveryRequired0 / 1

Wikipedia / Wikidata entity presenceRecommended2 / 4

Developer resource discoverabilityRecommended4 / 4

Brand name discoverabilityRequired3 / 3

robots.txt AI crawler policyRequired1 / 1

NPM/PyPI SDK packageRecommended1 / 1

Homepage not hidden from searchRecommended1 / 1

\+ 8 checks not in the score

AI Catalog discoveryEmergingNot counted

ARD entry validityRecommendedNot counted

ARD trust manifestEmergingNot counted

Listed in MCP registriesRecommendedNot counted

Agent platform configsEmergingNot counted

Agent Plugins manifestEmergingNot counted

Registry brandingEmergingNot counted

ChatGPT app listedRecommendedNot counted

Access21 of 302 to fix2 partly17 passed

Agent instruction / when-to-useRequired0 / 1

OpenAPI spec publishedRequired0 / 8

Content without JavaScriptRequired1.7 / 5

Trust anchor pagesRequired0.5 / 1

Sitemap existsRequired1 / 1

Not blocked by bot detectionRequired2 / 2

robots.txt agent-user policyRequired1 / 1

llms.txt existsRequired1 / 1

llms.txt formattingRecommended2 / 2

JSON-LD structured dataRequired2 / 2

Pricing info accessibleRequired3 / 3

Public API/docs linked from homepageRequired4 / 4

JSON-LD entity linking (sameAs)Recommended1 / 1

Metadata completenessRequired1 / 1

Organization schema completenessRecommended1 / 1

llms.txt links resolveRecommended2 / 2

Redirect hygieneRecommended1 / 1

Content behind authRecommended1 / 1

Developer portalRecommended5 / 5

Markdown content negotiation (acceptmarkdown.com)Recommended1 / 1

Agent crawler reachabilityRecommended1 / 1

\+ 19 checks not in the score

Agent discovery fileEmergingNot counted

Agent Skills index conformance (v0.2.0)EmergingNot counted

A2A / agent-cardEmergingNot counted

Listed on skills.shEmergingNot counted

pricing.md existsEmergingNot counted

NLWeb Schema FeedsEmergingNot counted

MCP well-known discoveryEmergingNot counted

HTTP Link headers (RFC 8288)EmergingNot counted

Markdown URL fallbackEmergingNot counted

Modular llms.txt per product areaEmergingNot counted

Skills.sh skill qualityEmergingNot counted

Markdown alternate linkEmergingNot counted

Markdown frontmatter metadataEmergingNot counted

Page token budgetRecommendedNot counted

Code fence validityEmergingNot counted

API catalog (RFC 9727)EmergingNot counted

Markdown agent docsEmergingNot counted

Bot-UA markdown servingEmergingNot counted

Content starts near the topEmergingNot counted

Usability7 of 4011 to fix2 partly1 passed

OAuth 2.0 supportRequired0 / 5

Scoped permissionsRecommended0 / 5

MCP server / manifestRequired0 / 10

Rate limit response headersRecommended0 / 2

Idempotency-Key supportRecommended0 / 3

JSON error responsesRequired0 / 3

REST typed error modelRecommended0 / 3

REST async-job patternRecommended0 / 2

CLI tool availableRecommended0 / 3

API schema complexity analysisRecommended0 / 2

Function calling compatibilityRecommended0 / 2

Agent onboarding frictionRecommended1 / 2

Agent-friendly 404sRecommended1 / 2

Public API with reachable endpointsRequired8 / 8

\+ 65 checks not in the score

MCP tool descriptionsRecommendedNot counted

MCP parameter schemasRecommendedNot counted

MCP server identityRecommendedNot counted

MCP tool listingRecommendedNot counted

MCP tool namingRecommendedNot counted

MCP auth mechanismRecommendedNot counted

MCP OAuth metadataRecommendedNot counted

MCP PKCE S256 supportRecommendedNot counted

Web Bot Auth directoryEmergingNot counted

OAuth Protected Resource metadata (RFC 9728)RequiredNot counted

auth.md existsEmergingNot counted

auth.md structureEmergingNot counted

auth.md walkthrough simulationEmergingNot counted

Agent auth discovery metadataEmergingNot counted

Agent auth WWW-Authenticate hintEmergingNot counted

agent\_auth endpoints reachableEmergingNot counted

MCP error handlingRecommendedNot counted

MCP modern transportRecommendedNot counted

WebMCP supportRequiredNot counted

REST versioning / deprecation policyRecommendedNot counted

REST pagination patternRecommendedNot counted

NLWeb /ask endpointEmergingNot counted

NLWeb streaming supportEmergingNot counted

REST response schema coverageRecommendedNot counted

MCP tool annotationsRecommendedNot counted

MCP server-card.jsonEmergingNot counted

Product + docs MCP coverageRecommendedNot counted

Sandbox / test environmentRecommendedNot counted

MCP resources exposedRecommendedNot counted

GraphQL typed error modelRecommendedNot counted

GraphQL versioning / deprecation policyRecommendedNot counted

GraphQL pagination patternRecommendedNot counted

GraphQL async-job patternRecommendedNot counted

GraphQL schema description coverageRecommendedNot counted

Accessible document structureEmergingNot counted

Native interactive controlsEmergingNot counted

Accessible names on controlsEmergingNot counted

Form control labelingEmergingNot counted

Accessibility-tree injection safety (bonus)EmergingNot counted

MCP Apps supportRecommendedNot counted

A2UI / generative UI supportEmergingNot counted

MCP Apps UI qualityRecommendedNot counted

MCP App view reachableRecommendedNot counted

MCP App view CSPRecommendedNot counted

MCP resource qualityRecommendedNot counted

MCP server instructionsRecommendedNot counted

MCP tools are distinguishableRecommendedNot counted

MCP tool output schemasRecommendedNot counted

MCP Apps checkout viewEmergingNot counted

WebMCP schema validityRecommendedNot counted

WebMCP schema detailRecommendedNot counted

WebMCP real toolsRecommendedNot counted

WebMCP entry pointRecommendedNot counted

WebMCP tool namesRecommendedNot counted

WebMCP read-only hintsRecommendedNot counted

WebMCP honest hintsRecommendedNot counted

WebMCP untrusted content hintRecommendedNot counted

WebMCP consequential hintRecommendedNot counted

WebMCP coverageRecommendedNot counted

Personal Agent Protocol discovery fileEmergingNot counted

PAP issuer bindingEmergingNot counted

PAP sign-in and extension configEmergingNot counted

PAP API and agent entriesEmergingNot counted

PAP endpoints reject unauthenticated agentsEmergingNot counted

PAP agrees with ARD and auth.mdEmergingNot counted

PaymentsNot scored yet

\+ 5 checks not in the score

MPP payment protocolRequiredNot counted

x402 payment protocolRequiredNot counted

UCP - Universal Commerce ProtocolRequiredNot counted

ACP - Agentic Commerce ProtocolRequiredNot counted

Agent wallet checkoutRequiredNot counted
