---
title: "What the Agent Readiness Leaderboard Shows Is Failing"
description: "A read of the weekly public leaderboard: which of find, read, use, and pay fails most often, and what that says about the sites agents actually reach."
canonical: "https://www.agentrail.co/blog/state-of-agent-readiness"
author: "Vivek Mittal"
published: "2026-11-02T00:00:00Z"
last_updated: "2026-11-02T06:30:30.877Z"
---

# What the Agent Readiness Leaderboard Shows Is Failing

A read of the weekly public leaderboard: which of find, read, use, and pay fails most often, and what that says about the sites agents actually reach.

![White trophy on a three-step podium against a dark green gradient](https://images.ctfassets.net/1htfyfnn1o5y/3Zb6IfBkwM5a7fMXBJoSxN/0deb2325e9a5a8e168e044c1c04086bb/a2c145b8-7ea2-421d-87a9-d544d802a493.png?w=1600&fm=webp&q=72)

The [agent readiness leaderboard](https://www.agentrail.co/agent-readiness) is a weekly public scan of whether an agent can find, read, use, and pay for a product. On the November 1, 2026 snapshot it holds 54 companies, averages 64 out of 100, and tops out at 93 for agent-qa and Railway. The grade mix is five A, sixteen B, twenty-two C, ten D, and one F.

This piece is a field read of that board. It asks which of the four questions fails most often in public reports, and what that implies for sites agents actually reach. It is not a definition of [agent readiness](https://www.agentrail.co/blog/what-is-agent-readiness). It is not a walkthrough of the checker UI. Those jobs live elsewhere. The patterns below come from the public board and from cited markdown reports, not from private journey data.

> **Key Takeaways**
>
> • On the November 1, 2026 board, most companies sit at grade C or below. An average of 64 is the middle of a hard public scan, not a soft marketing score.
>
> • In a sample of 20 public reports, discovery usually sits near its 20-point cap. Access and usability take more of the remaining points.
>
> • Normalized to each pillar's maximum, that sample fills about 89% of Discovery, 75% of Access, and 61% of Usability. Payments is not scored yet.
>
> • High discovery with weak access or usability is common: LangChain and GitHub show the shape clearly.
>
> • A readiness grade is a precondition. Whether an agent finished the task is [agent experience](https://www.agentrail.co/blog/agent-experience), a different record.

## The board is one public scan, not a finished task

Every company on the board gets the same public scan. The scan uses what an agent can reach without a login and without a private API. The four questions behind the score are find, read, use, and pay. On the board those map to Discovery (up to 20), Access (up to 30), Usability (up to 40), and a Payments bonus that does not yet drive the graded columns the way the first three do.

That design matters for how you read a failure. A miss means the public contract was incomplete under this instrument. It does not mean a scripted eval failed in your lab, and it does not mean a production task completed or died. The definition of the property is in [what agent readiness is](https://www.agentrail.co/blog/what-is-agent-readiness). How the number is assembled for a check is on the [agent readiness checker](https://www.agentrail.co/agent-readiness/checker). This page stays with what the published board is showing.

## Most of the board sits at C or below

Thirty-three of the fifty-four companies on the November 1, 2026 snapshot are grade C, D, or F. Sixteen are B. Five are A. The average score is 64. The top score is 93, shared by agent-qa (vostride.com, community) and Railway.

Those letters are words for a public result. The board describes A as strong, B as good, C as fair, and D or F as hard for an agent to find, read, or use. It does not publish numeric cutoffs for the letters, and this article will not invent them.

The shape is still clear. Presence on the board is not the same as being ready. The center of mass is fair, not strong. If your mental model of "agent-ready companies" is a short list of A grades, the public list is longer and harder.

## Discovery is usually the strongest pillar in the sample

In the public markdown reports cited below, Discovery often sits near 20 even when the overall score is mid-pack. LangChain scores 19 out of 20 on Discovery with a 47 overall. GitHub scores 19 out of 20 with a 53. Stripe scores 19 out of 20 with a 79. ClickHouse, Airbyte, and Railway sit at a full 20 out of 20.

Discovery is not free for everyone. Fly.io scores 15 out of 20 while still ranking near the top overall. Modal and Daytona sit at 12 out of 20 with overall scores of 51 and 50. The pattern is still directional: among the cited reports, the find question is the pillar least likely to be empty.

An outside instrument points the same way for different checks. Cloudflare's April 17, 2026 Agent Readiness write-up looked at a filtered set of highly visited domains and found `robots.txt` on 78% of sites, Content Signals on 4%, markdown content negotiation on 3.9%, and MCP Server Cards plus API Catalogs together on fewer than 15 sites. Crawler basics are common. Agent-specific contracts are rare. Cloudflare's score is not Agentrail's score. Treat it as industry context, not a merged average.

## Access and usability are where the board loses points

The useful comparison is not raw points. Discovery is capped at 20, Access at 30, Usability at 40. To see which question fails hardest, normalize each pillar to its own maximum.

On November 1, 2026 we fetched twenty public markdown reports from `https://www.agentrail.co/markdown/agent-readiness/{domain}`: railway.com, vercel.com, fly.io, stripe.com, langchain.com, github.com, cloudflare.com, render.com, digitalocean.com, mongodb.com, modal.com, daytona.io, clickhouse.com, duckdb.org, airbyte.com, motherduck.com, neo4j.com, databricks.com, mastra.ai, and attio.com. Mean fill rates in that sample were about 89% of Discovery's 20 points, 75% of Access's 30, and 61% of Usability's 40. Payments read "not scored yet" on every report in the set.

That is not a board-wide failure percentage for all 54 companies. It is a method you can redo: open the same public reports, average the three pillars, and see whether usability still sits lowest. In this sample it does.

The report chrome matches the arithmetic. Access findings often name content that disappears without JavaScript, and a missing agent discovery file. Usability findings often name agent onboarding friction, missing or thin `auth.md`, OAuth support, scoped permissions, and Web Bot Auth directory gaps. Top sites still miss some of those. Railway's public report lists usability fixes even at 37 out of 40. Vercel lists auth.md and onboarding friction while sitting at 35 out of 40.

Contrast pairs make the ownership clear. Vercel (19 / 29 / 35) and Railway (20 / 27 / 37) keep access and usability high with discovery. LangChain (19 / 12 / 12) and GitHub (19 / 14 / 16) keep discovery high and lose the score on access and usability. An agent can often find those products. The public scan says it cannot reliably read and use them the way this rubric defines use.

## Findable is not usable

| Company | Score | Discovery /20 | Access /30 | Usability /40 | Payments |
| --- | --- | --- | --- | --- | --- |
| Railway | 93 | 20 | 27 | 37 | not scored yet |
| Vercel | 92 | 19 | 29 | 35 | not scored yet |
| Fly.io | 91 | 15 | 29 | 38 | not scored yet |
| Stripe | 79 | 19 | 22 | 30 | not scored yet |
| GitHub | 53 | 19 | 14 | 16 | not scored yet |
| LangChain | 47 | 19 | 12 | 12 | not scored yet |

Sources: public markdown reports for each domain on agentrail.co, retrieved November 1, 2026.

The table is the argument in short form. Four of the six rows clear most of Discovery. Only the top three clear most of Usability. If your team equates "agents know who we are" with "agents can operate us," the board disagrees.

A readiness grade still does not prove a task finished. It records the public precondition. What happens when an agent tries to complete work on your product is [agent experience](https://www.agentrail.co/blog/agent-experience). Keep the two measurements separate so a green discovery pillar does not get mistaken for a finished install.

## Payments is a blank column, not a clean bill of health

Across the cited reports, Payments shows as "not scored yet." Several reports note partial payment checks that will count when scoring opens. Until then, the graded story on the board is find, read, and use.

Do not read the blank as proof that agents can pay. Do not invent a board-wide pay-failure rate either. The honest state of pay on this leaderboard is that the bonus column is not yet driving the grades the way Discovery, Access, and Usability do. When that changes, the failure pattern may shift. Recheck the board before you cite a pay figure.

## What to do with a mid-board score

Start from the pillar furthest from its cap. If Discovery is 19 and Usability is 12, the public contract problem is not "agents cannot find us." It is "agents cannot take a supported action under this scan." Open the company's public report for the evidence lines and fix prompts. Then decide whether the next measurement is another readiness scan, or a live task you reconstruct after an agent arrives.

The [leaderboard](https://www.agentrail.co/agent-readiness) is the shared view. The checker link above is enough for the score itself. Readiness scans are included with Agentrail. Journeys and simulations are priced on their own on the [pricing](https://www.agentrail.co/pricing) page.

## Frequently asked questions

### Which pillar fails most on the public board?

In the twenty public reports cited here, Usability has the lowest mean fill rate as a share of its maximum, then Access, then Discovery. That is a sample method, not a census of all fifty-four companies. Payments is not scored yet, so it is not in the comparison.

### Does a high discovery score mean agents succeed?

No. Discovery asks whether an agent can find the product under this scan. Access and usability ask whether it can read and act. LangChain and GitHub both show high discovery with weak access and usability on their public reports.

### Are payments included in the grade?

Payments is a bonus pillar. On the cited November 1, 2026 reports it still reads "not scored yet." Do not treat that blank as a pass or a fail rate.

### How often does the board update?

Every company on the board is rescanned every week on the same checks. The date on the leaderboard page is the latest of those scans. Company report pages also show when a path or score was last run.

## Read the pillar that is empty

The state of agent readiness on this public board is not "nobody can be found." It is closer to "many products are findable, fewer are readable and usable under one shared scan, and pay is not scored yet." Open the [agent readiness leaderboard](https://www.agentrail.co/agent-readiness), pick the pillar that is furthest from its cap, and treat that as the first public contract to repair. Then measure the encounter itself with [agent experience](https://www.agentrail.co/blog/agent-experience) once an agent actually arrives.