# Glintbase — Agent Readiness Infrastructure > Glintbase measures how well AI agents (Claude Code, Cursor, ChatGPT, Perplexity and others) can discover, read and operate a product's documentation, APIs and MCP servers, and shows teams what to fix. The measurement is the Agent Readiness Score (ARS 3.3): 137 registered checks across four layers (Discovery, Access, Usability, Payments), with points awarded only for evidence the scanner actually observes. HTML version: https://glintbase.dev/ ## Products - **Hosted scanner** — https://scan.glintbase.dev — free, no sign-up. Scan any public URL and get an ARS 3.3 score, per-check evidence and fixes. - **CLI** — `npx @glintbase/cli audit ` (@glintbase/cli 3.3.0). Commands include audit, check, ci, simulate, fix, mcp and mcp lint. Scores with the same engine as the hosted scanner (@glintbase/scanner-core). - **MCP servers** — hosted, zero-auth: https://glintbase.dev/api/mcp (proxy of https://scan.glintbase.dev/api/mcp; 5 tools). Local: `glintbase mcp` from @glintbase/cli (17 tools) or `npx -y @glintbase/mcp` (12 scanner tools, stdio). - **GitHub Action / CI gate** — `glintbase ci --fail-under 80` fails a build when the score drops below a threshold. ## Latest release **v3.5.1 — ARS 3.3.1: The Scanner Declares Itself** (October 2026). The scanner now identifies itself as Mozilla/5.0 (compatible; GlintbaseScanner/3.5; +https://glintbase.dev/research/bot) instead of sending a desktop Chrome user agent. Before switching we scanned every leaderboard company as both identities from Vercel's network: the honest identity was never blocked where Chrome was served. Scores from ars-3.3.0 are not comparable with ars-3.3.1; the leaderboard was rescanned on release. Full changelog: https://glintbase.dev/changelog.md ## Research - [The Silent Wall: How Bot Defences Treat a Self-Declared AI Agent](https://glintbase.dev/research/the-silent-wall.md) (Oct 13, 2026) — 150 popular sites fetched as a declared research agent, an unlabelled script and a headless browser: 16.0% vs 18.7% vs 18.7% of homepages challenged or blocked — no significant difference. robots.txt targets AI training (23.6%) far more than user-initiated agents (12.1%). - [Remote MCP Server Reliability: A Census of the Official MCP Registry](https://glintbase.dev/research/remote-mcp-server-reliability.md) (Oct 13, 2026) — Protocol-only probe of 3,000 of 26,365 remote endpoints: 56.9% work anonymously, 82.7% of auth-gated endpoints give a followable challenge, 13.3% speak MCP 2026-07-28, and 24.9% fail to report errors in-channel. - [The Glintbase Integration Index: Cloud Sandboxes for Autonomous AI Coding Agents](https://glintbase.dev/research/sandboxes-edition-1.md) (Oct 7, 2026) — Pilot benchmark evaluating Modal, E2B, boat.dev, Vercel Sandbox, and Daytona across 119 front-door scans and 30 live agent trials (28/30 passed). Corrected 2026-10-12: prompts differed by provider, so the Integration Tax ranking is indicative only. - [State of Agent Readiness 2026: Global Benchmark Report on Developer Infrastructure](https://glintbase.dev/research/state-of-agent-readiness-2026.md) (Aug 5, 2026) — Evaluation of 100 developer platforms across 1,000 deterministic, simulated Pathfinder traversal scans (mean ARS 50.7, median 52, max 79). Corrected 2026-10-12: the data, not the PDF, is the source of truth. - Research hub: https://glintbase.dev/research.md - GlintbaseResearch crawler (what it requests, how to opt out): https://glintbase.dev/research/bot.md ## Latest articles - [ARS 3.3: Gates, Dual-Era MCP and What Changed on the Leaderboard](https://glintbase.dev/blog/ars-3-3-release.md) (2026-10-24) — ARS 3.3 caps the score of sites agents can't read, stops scoring llms.txt for merely existing, probes MCP servers in the 2026-07-28 protocol and finds Server Cards where the final spec puts them. We scanned all 103 leaderboard companies with the release code: the average fell from 51.5 to 49.1, and 28 sites hit a gate. - [What the llms.txt Evidence Actually Says](https://glintbase.dev/blog/what-the-llms-txt-evidence-says.md) (2026-10-24) — Three large studies found no link between serving llms.txt and AI citations, and most llms.txt files are never requested. Yet agents that find one through a link do use it. We read the evidence, checked 102 developer companies ourselves, and changed how the Agent Readiness Score treats the file. - [Agent-Ready Means Governed, Not Just Reachable](https://glintbase.dev/blog/agent-ready-means-governed.md) (2026-10-23) — Two large October 2026 surveys found that most enterprises already run AI agents but few can trace, see or stop them, and a third have found MCP servers nobody approved. Readiness has a second half: agents that can be identified, scoped and stopped. Here is what of that a site can show from the outside, and what it can't. - [How ARS Compares With AgentReady, Ora and the IETF Draft](https://glintbase.dev/blog/how-ars-compares-agentready-ora-ietf.md) (2026-10-23) — Three public definitions of agent readiness appeared in October 2026. We mapped every Agent Readiness Score check against AgentReady.org's requirements, Ora's four layers and the IETF individual draft, and explain where our weights come from, including what we borrowed and what we still need to prove. - [The MCP Auth Fix That Ships Turned Off](https://glintbase.dev/blog/mcp-auth-fix-ships-turned-off.md) (2026-10-23) — Between 28 September and 15 October 2026 the official MCP SDKs published 13 security advisories. The fix for servers accepting tokens issued for another service is opt-in in both the Python and TypeScript SDKs: upgrading alone changes nothing. Here is what to upgrade to and the two changes that actually turn the check on. - [What 3,000 MCP Servers Told Us About the Basics](https://glintbase.dev/blog/what-3000-mcp-servers-told-us.md) (2026-10-23) — We probed a random sample of 3,000 remote endpoints from the official MCP Registry using only protocol requests. Most work, auth is mostly done right, tool schemas are strong, but one in four servers can't tell an agent that a call failed, and only 13% speak the current protocol revision. - [Why Agent-Readiness Scores Disagree](https://glintbase.dev/blog/why-agent-readiness-scores-disagree.md) (2026-10-23) — We ran three public sites through four agent-readiness scanners, our own included. The same site scored anywhere from 47 to 96 depending on the tool, and every tool ranked the three sites differently. Here is what each one measures, and how to read any readiness score. - [Your robots.txt Lets AI Agents In. Your Edge Might Not](https://glintbase.dev/blog/robots-txt-said-yes-the-edge-said-no.md) (2026-10-23) — We fetched the same public pages on 150 popular websites as a declared AI agent, an unlabelled script and a headless browser. Saying we were an agent cost nothing, yet on 16.8% of sites whose robots.txt allowed us the edge still stopped us. Here is what we found, and what site owners should check. All articles: https://glintbase.dev/blog.md ## Machine-readable surfaces - llms.txt: https://glintbase.dev/llms.txt - llms-full.txt: https://glintbase.dev/llms-full.txt - Documentation (Markdown): https://glintbase.dev/docs.md - ARS 3.3 methodology (Markdown): https://glintbase.dev/methodology.md - OpenAPI 3.1: https://glintbase.dev/openapi.json - MCP manifest: https://glintbase.dev/.well-known/mcp/manifest.json - Agent Resource Discovery: https://glintbase.dev/.well-known/ard.json - Agent instructions: https://glintbase.dev/agents.md - Authentication (none required): https://glintbase.dev/auth.md - Sitemap: https://glintbase.dev/sitemap.xml Every page listed in the sitemap that has a Markdown twin also returns Markdown when requested with `Accept: text/markdown`. ## Contact contact@glintbase.dev · https://glintbase.dev/support