AI-agent-readiness scanners have arrived. In mid-2026 you have real choice about which one to run against your site. Two of the most active in the space are AgentGrade and AgentFix. They look similar on the surface (paste a URL, get a report on how AI agents see your site) but they solve different problems. This guide is an honest side-by-side so you pick the right tool for what you actually need.
Short version: AgentGrade grades your site across the widest protocol matrix on the market today. AgentFix scans a similar signal set and then ships you the missing files so you can stop reading specs. If you want a letter grade, use AgentGrade. If you want the fix, use AgentFix.
What each tool actually does
AgentGrade
AgentGrade is a free scoring service that assigns a letter grade to any URL based on six categories of agent-facing signals: payment protocols (x402, MPP, L402, SPT), Model Context Protocol endpoints, discovery files (llms.txt, agents.txt, A2A), decentralised identity (WebFinger, DID, Nostr), and general infrastructure. It also exposes its own MCP endpoint so other agents can query grades programmatically. Paid tiers are marked as "planned" but not live at time of writing.
The strength is breadth of protocol coverage. If a spec exists in the agent-web ecosystem, AgentGrade probably checks for it. The weakness is that the output is designed for agents to consume, not humans to act on. You get a letter and a list of missing pieces. What you do with that list is up to you.
AgentFix
AgentFix scans 33 signals (a subset heavier on the discovery, schema, and protocol layers, lighter on the emerging DID/payment layers) and then generates a personalised ZIP fix-pack containing the files your site is missing. Real llms.txt written from your content, schema.org JSON-LD tuned to your business, /.well-known/agent-card.json and /.well-known/mcp/server-card.json declaring your skills, plus platform-specific install guides for WordPress, Webflow, Shopify, Tilda, and cPanel.
The strength is remediation. You are not left holding a checklist. The weakness is that the OSS mini scanner covers 12 of the 33 signals; the full 33 land only when you buy a pack. A free CLI is available for the 12 that matter most for discovery.
Side-by-side matrix
| Dimension | AgentGrade | AgentFix |
|---|---|---|
| Signals checked | Widest matrix in the space (6 categories including payments and DID) | 33 signals focused on discovery, schema, agent protocols |
| Output format | Letter grade + category breakdown + MCP endpoint for agents | Score + human-readable report + ZIP with the files to install |
| Price | Free today; paid "planned" | Free scan; $1 Mini pack, $29 full pack, $99 pack + drift monitoring |
| Model | Report-first | Remediation-first (files, not just findings) |
| Best for | Understanding where every emerging protocol sits | Getting the site fixed today |
| Machine access | MCP endpoint (agents can pull grades directly) | Public OpenAPI, plus a webhook when your pack is ready |
| Free open-source companion | Not published | agentfix-mini-scanner on GitHub (12 signals, MIT, npx) |
Where AgentGrade wins
- Payment protocol coverage. AgentGrade actively checks for x402, MPP, L402, and SPT. If your site sells anything and you want to be ready for agent-driven commerce, that visibility matters.
- Decentralised identity. WebFinger, DID, Nostr signals are surfaced. Most sites do not ship them yet, but if you do, AgentGrade gives you credit.
- Agents can query it. The MCP endpoint means other agents can look up your grade without a browser. Useful if you want your grade to be discoverable programmatically.
- Free. No paywall today.
Where AgentFix wins
- You get the files, not the homework. The pack contains a real
llms.txtwritten from your content, real JSON-LD tuned to your business, and the two.well-knownmanifests. No copy-pasting from spec pages. - Install guide per platform. WordPress, Webflow, Shopify, Tilda, cPanel. The nine screens between "the file exists on my desktop" and "the file is live at
example.com/llms.txt" are documented for each. - One-time price, not subscription. $1 for the Mini (just
llms.txt), $29 for the full 33-signal pack, $99 if you also want monthly drift monitoring. No SaaS lock-in. - Free CLI on GitHub. Run
npx agentfix-mini-scanner example.comin your terminal. 12 checks, zero dependencies, MIT licensed. Fits into CI. - Free interactive visualisation. The AgentFix Workspace iceberg shows what Google sees versus what agents reach for versus what the ideal shape looks like, side by side.
The one-line difference
Scanners tell you your site fails. AgentFix gives you the exact files that make it pass.
Both are useful. If you are a spec nerd who wants to see the widest possible signal matrix and understand exactly what agent-web protocols exist in 2026, AgentGrade is the map. If you are a site owner who needs the score up before your next AI review, AgentFix is the toolbox.
Which one should you actually run today?
Run AgentGrade if
- You are building an agent-native business and need to know which emerging protocols you should adopt next.
- You already have engineers who can implement anything you point them at, you just need the map.
- You want your grade to be queryable via MCP so other agents can find it.
Run AgentFix if
- You run a business site (WordPress, Webflow, Shopify, Tilda) and you do not want to become an agent-protocol expert to serve one.
- You want the files installed by end of week, not next quarter.
- You want to spend $1 to $99 once instead of monthly forever.
- You want a CLI you can run in CI so a bad deploy does not silently break your llms.txt.
Run both if
- You are auditing a client site and want two independent opinions before quoting the work.
- You are writing a report on the state of agent readiness in your industry and need the widest signal coverage plus a working remediation pipeline.
FAQ
Is AgentGrade the same as an AgentFix competitor?
Adjacent, not identical. Both score sites for AI-agent readiness. AgentGrade stops at the score and offers a machine-readable MCP endpoint. AgentFix continues past the score into personalised remediation files.
Can I use AgentGrade's report to buy an AgentFix pack?
Yes. If AgentGrade shows you missing llms.txt, missing schema, or missing agent-card, the AgentFix pack ships all three (plus the other 30 signals) as installable files. You do not lose any information by moving from one to the other.
Do the tools agree on scores?
Mostly. Both check llms.txt, schema.org, robots.txt for AI crawlers, and the .well-known manifest set. AgentGrade weighs payment and DID signals more heavily. AgentFix weighs schema completeness and per-page hygiene (heading hierarchy, image alt, meta description length) more heavily. So a site strong on protocols but weak on schema will get a higher grade from AgentGrade than from AgentFix, and vice versa.
What about the free open-source scanner?
agentfix-mini-scanner is the MIT-licensed CLI that runs 12 of the 33 AgentFix signals. Zero dependencies, Node 18+. Run it via npx agentfix-mini-scanner your-site.com. If you want the remaining 21 signals plus the ZIP fix pack, that lives behind the one-time pack purchase at agentfix.pro.
Which tool checks Model Context Protocol servers correctly?
Both look for /.well-known/mcp/server-card.json. AgentGrade additionally validates the linked MCP endpoint returns a well-formed manifest. AgentFix validates the manifest structure and confirms your declared tools match your OpenAPI where present. Different depths, both useful.
The take-away
Do not treat this as a shootout. AgentGrade is the widest map of agent-web protocols in 2026. AgentFix is the fastest path from "my score is low" to "my score is high" for a real business site. Grade first with AgentGrade if you are a protocol tourist. Fix first with AgentFix if you are a business owner. Grade and fix together if you are a consultant billing hours to make it all happen.
Ready to see where your site actually sits? Run the free AgentFix scan or grab the MIT CLI on GitHub. If your score is low, the pack starts at $1.