Changelog

What's new at Glintbase

Releases and improvements across the Agent Readiness Scanner, MCP server, CLI, and agent skill. Updated as we ship.

  1. Added

    Shareable scorecards & competitive context

    • —Every scan report now generates a shareable 1200×630 scorecard image for X, LinkedIn, or direct download.
    • —Scan reports show competitive context — how your score compares against well-known products on the leaderboard.
  2. Added

    Public leaderboard

    • —Launched the Agent Readiness leaderboard at scan.glintbase.dev — every public scan is ranked by ARS score.
  3. Added

    Agent skill for coding agents

    • —Published an installable agent skill so coding agents (Claude Code, Cursor, and others) can run readiness scans and apply remediation autonomously.
    • —Skill source lives in the open at github.com/glintbase/glintscanner/tree/main/skills.
  4. Added

    @glintbase/cli

    • —Released the Scanner CLI: run `npx @glintbase/cli scan <url>` locally or in CI.
    • —Supports --fail-under thresholds for gating deploys on agent readiness.
  5. Added

    @glintbase/mcp — Model Context Protocol server

    • —Released the MCP server exposing the full scan pipeline as 9 structured tools: discover_surfaces, check_reachability, parse_spec, crawl_pages, deep_crawl, build_knowledge_graph, run_journeys, score_readiness, get_remediation.
    • —Runs locally via `npx -y @glintbase/mcp` — no API keys required for core tools.
  6. Improved

    deep_crawl extraction engine

    • —New deep crawl profile recovers real page content from JS-rendered docs sites (Next.js, Docusaurus, Nextra, SPA shells) via embedded data extraction — __NEXT_DATA__, RSC flight chunks, JSON-LD, and noscript fallbacks.
    • —Pages that previously scored as thin shells are now evaluated on their true content.
  7. Changed

    ARS 1.0 scoring model

    • —The Agent Readiness Score is now a versioned, weighted composite across eight dimensions: Discoverability, Machine Entrypoints, Canonical Sources, Content Quality, Graph Connectivity, Journey Success, Freshness, and Runtime Validity.
    • —Anti-gaming checks verify that published surfaces (llms.txt, OpenAPI, MCP configs) are real and reachable, not decorative.
  8. Changed

    Agent Readiness Scanner v2

    • —Rebuilt the scanner around a deterministic pipeline: surface discovery → crawl → knowledge graph → agent journey simulation → scoring → remediation.
    • —Agent journeys test whether an AI agent can complete real integration tasks using only your documentation.