@glintbase/mcp

The Glintbase MCP

Zero-config agent-readiness tools for AI coding agents. No API keys, no .env — your agent calls the tools and reasons over the results with its own subscription.

.mcp.jsonjson

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

Zero-key deep_crawl

JS-rendered docs (Next.js, Docusaurus, SPAs) return an empty shell over raw HTTP. The deep_crawl tool recovers the real content from embedded framework payloads (__NEXT_DATA__, RSC flight chunks, JSON-LD, <noscript>) — no API key.

Optional Firecrawl

For the hardest SPAs, add "env": { "FIRECRAWL_API_KEY": "fc-..." } to the config and pass profile: "deep" for premium extraction. Not required.

The golden path

Agent tool sequencetext

discover_surfaces  ->  deep_crawl  ->  score_readiness  ->  get_remediation

Run deep_crawl before score_readiness for JS-heavy sites — it populates the session so scoring reuses the recovered pages instead of re-crawling a thin shell.

9 composable tools

discover_surfacesFind all machine-readable entrypoints (llms.txt, OpenAPI, MCP, docs, ...)

check_reachabilityQuick single-URL reachability + soft-404 detection

parse_specParse OpenAPI / llms.txt / MCP configs without crawling

crawl_pagesBudgeted crawl with priority queue (quick/deep profiles)

deep_crawlIn-depth crawl that recovers JS-rendered content — no API key

build_knowledge_graphSemantic graph from crawled pages

run_journeysDeterministic agent journey simulations (no LLM)

score_readinessFull pipeline → composite ARS score (auto-runs missing stages)

get_remediationPrioritized fixes with templates and expected score impact

Hosted remote MCP

Coming soon

Connect your agent to Glintbase in one line — no install. A hosted Streamable-HTTP endpoint at scan.glintbase.dev/mcp, gated behind an API key and rate limits. Until then, the npx setup above is the way in.

Scan Your Ecosystem All tools