Glintbase Docs

The fastest way to get an Agent Readiness Score is the hosted scanner. Enter any product or documentation URL and the pipeline runs discovery, crawling, knowledge-graph construction, agent journey simulation, and scoring — no signup required to start.

Results include the composite ARS score (0–100), a per-dimension breakdown, agent journey traces, and prioritized remediation advice. Reports are shareable via a stable link and a scorecard image.

@glintbase/cli runs the same scan pipeline from your terminal. Use it in CI/CD to catch agent-readiness regressions before they ship.

# Scan any product URL and print the Agent Readiness Score
npx @glintbase/cli scan https://docs.example.com

# Gate CI/CD on a minimum score — exits non-zero below the threshold
npx @glintbase/cli scan https://docs.example.com --fail-under 70 --quiet

Source: github.com/glintbase/glintscanner/tree/main/cli

@glintbase/mcp exposes the scan pipeline as nine Model Context Protocol tools: discover_surfaces, check_reachability, parse_spec, crawl_pages, deep_crawl, build_knowledge_graph, run_journeys, score_readiness, and get_remediation. Add it to any MCP-capable client (Claude Desktop, Cursor, and others):

{
  "mcpServers": {
    "glintbase": {
      "command": "npx",
      "args": ["-y", "@glintbase/mcp"]
    }
  }
}

No API keys are required for the core tools. The machine-readable manifest for this server is published at glintbase.dev/mcp.json.

The Glintbase agent skill teaches coding agents to run readiness scans and apply remediation autonomously inside your repo.

# Clone the scanner repo and copy the skill into your agent's skill directory
git clone https://github.com/glintbase/glintscanner.git
cp -r glintscanner/skills/* ~/.claude/skills/

Skill source: github.com/glintbase/glintscanner/tree/main/skills

The Agent Readiness Score is a versioned, weighted composite of eight dimensions. Bands: 85+ AI-Native, 70–84 AI-Ready, 50–69 AI-Capable, below 50 AI-Limited.

Discoverability

Can agents find your product surfaces at all — robots policy, sitemaps, crawlable structure.

Machine Entrypoints

llms.txt, OpenAPI specs, MCP configs — verified real and reachable, not decorative.

Canonical Sources

A resolvable docs root, source repository, and package surfaces agents can treat as ground truth.

Content Quality

Substantive pages with real code examples, not thin marketing shells.

Graph Connectivity

How well your pages interlink — isolated content is invisible to traversing agents.

Journey Success

Simulated agent journeys: can an agent complete real integration tasks from your docs alone.

Freshness

Changelog and status signals that tell agents the product is alive and current.

Runtime Validity

Every advertised surface is probed live — dead links and 404s count against you.

Glintbase practices what it scans. These machine-readable surfaces are live on this domain:

• /llms.txtConcise index of concepts, products, and canonical links for LLMs.
• /llms-full.txtConsolidated full-text documentation for LLM ingestion.
• /mcp.jsonManifest for the @glintbase/mcp Model Context Protocol server.
• /sitemap.xmlFull sitemap of crawlable routes.
• /changelogProduct release history.
• /statusLive availability of glintbase.dev and scan.glintbase.dev.

Questions or issues? See /support.