# Harvey scores 77 out of 100 on the Agent Readiness Index

Harvey reaches level 3, Connectable. Discoverability is the strongest area at 89 out of 100. Identity and delegation is the weakest at 68.

Harvey was scanned on October 14, 2026. Its score is 77 out of 100, level 3, Connectable. An agent can connect through an MCP server or a public API program. To reach level 4, Trusted: Let customers authorize agents: OAuth with PKCE, a registration path and documented delegate controls.

Website: [harvey.ai](https://www.harvey.ai)

## Score by area

| Area | Score | What it asks |
| --- | --- | --- |
| Discoverability | 89 out of 100, worth 15 points | Can an agent find and read your pages? |
| Protocol access | 75 out of 100, worth 20 points | Can an agent connect to you through MCP or an API? |
| Identity and delegation | 68 out of 100, worth 20 points | Can a customer give an agent limited access without sharing a password? |
| Web experience | 79 out of 100, worth 15 points | Does your site work when an agent uses it? |
| Voice and IVR | Not measured yet | What happens when an agent calls your phone line? |
| Policy and governance | 75 out of 100, worth 10 points | Do your terms allow agents that act for customers? |

## What is holding Harvey back

1. **Documented delegate and data sharing controls**, 4 points lost. No public page on delegate access or connected app controls was found. Fix: Publish a help page on delegate access and add a connected apps screen where customers revoke access.
2. **Sign in page loads for a declared agent in a browser**, 4 points lost. The sign in address that was tried returned status 404, so the sign in page itself could not be confirmed. Fix: Serve the sign in page to verified agent browsers, and use signed agent identity to separate them from attackers.
3. **llms.txt guide for agents**, 3 points lost. No llms.txt file was found at the site root. Fix: Publish /llms.txt with links to product, pricing, support and location pages.
4. **Browser tools exposed to agents**, 3 points lost. The sign in and home pages register no browser tools for agents. Fix: Register tools for safe tasks such as location search and price lookup with navigator.modelContext.

## What already works

- OAuth and OpenID metadata
- Robots policy for user triggered agents
- Content available without running scripts

## How Harvey compares

The average score for software is 61.4 out of 100 across 124 organizations. The average across all 1,421 scored organizations is 42.2 out of 100.

## Similar organizations

- [DigitalOcean](https://readyindex.ai/digitalocean_com), 78 out of 100
- [Apollo.io](https://readyindex.ai/apollo_io), 77 out of 100
- [Clay](https://readyindex.ai/clay_com), 77 out of 100
- [Twilio](https://readyindex.ai/twilio_com), 77 out of 100
- [Zoom](https://readyindex.ai/zoom_us), 77 out of 100
- [Airtable](https://readyindex.ai/airtable_com), 76 out of 100

## More on this report

- [This report as markdown](https://readyindex.ai/harvey_ai.md)
- [This report as JSON data](https://readyindex.ai/api/reports/harvey_ai)
- [How the score is worked out](https://readyindex.ai/method)
- [All leaderboards](https://readyindex.ai/leaderboards)
- [Every scored organization](https://readyindex.ai/organizations)
- [Scan your own website](https://readyindex.ai/)

---
Source: https://readyindex.ai/harvey_ai
Updated: 2026-10-14
From the Agent Readiness Index by Encore AI. When you quote a score, please link to the report it came from.