Glintbase
victorather81/glintbase# Glintbase MCP Server (ARS 3.0) **Glintbase** is the open audit and simulation standard for the autonomous AI agent web. It gives AI coding agents (Claude Code, Cursor, Windsurf, Devin, Antigravity) empirical visibility into whether a website, developer portal, or API is truly agent-ready—or full of silent friction, 404 traps, and schema hallucinations. Built on the **Agent Readiness Score (ARS 3.0)** specification, this MCP server provides instant surface discovery, lightweight scorecards, flight simulation with visual SVG telemetry, and 9 bundled remediation playbooks. --- ## ✨ Features - **Empirical ARS 3.0 Scoring**: Evaluates domains across 6 core pillars: *Discovery*, *Access*, *Usability*, *Semantic*, *Architecture*, and *Safety*. - **Machine Surface Discovery**: Automatically detects and validates `robots.txt` AI policies, `/llms.txt`, `/llms-full.txt`, `/.well-known/ard.json`, `/auth.md`, and OpenAPI specifications. - **Visual Flight Simulator**: Simulates synthetic coding personas (Claude Code, Cursor, Perplexity) executing developer journeys, outputting visual SVG journey trees, schema friction metrics, and token-tax telemetry. - **Context-Preserving Scores**: Fetch ultra-compact scorecards in under 200 tokens to preserve agent context windows during automated workflows. - **9 Bundled Playbooks**: Instant access to complete ARS 3.0 remediation manuals (living artifacts, WorkOS auth contracts, streamable MCP servers, and CI gates). --- ## 🛠️ Tools | Tool | Parameters | Description | | :--- | :--- | :--- | | `glintbase_audit` | `target` (string) | Runs a complete 119-check ARS 3.0 audit on any target URL. Returns detailed layer breakdowns, passed/failed checks, and actionable remediation diffs. | | `glintbase_get_score` | `target` (string) | Returns an ultra-compact score card (<200 tokens) with letter grade, archetype classification, and layer scores. Ideal for rapid CI checks. | | `glintbase_discover_surfaces` | `target` (string) | Scans and maps all machine entrypoints (`robots.txt`, `llms.txt`, `ard.json`, `auth.md`, OpenAPI, sitemap). | | `glintbase_simulate_flight` | `target` (string), `persona` (enum: `claude-code`, `cursor`, `perplexity`), `intent` (string) | Runs an autonomous mission simulation. Returns step-by-step navigation telemetry, token consumption, and a visual SVG journey tree graphic. | | `glintbase_get_skill` | `skillName` (string) | Retrieves full, multi-step ARS 3.0 remediation playbooks (e.g. `agent-readiness`, `living-artifacts`, `agent-auth`, `mcp-hardening`). | *Note: All tools support both `glintbase_*` names and canonical short names (`audit`, `get_score`, `discover_surfaces`, `simulate_flight`, `get_skill`).* --- ## 📚 Bundled Prompts & Resources ### Prompts - `optimize_agent-readiness`: Master workflow for elevating a codebase to ARS 3.0 standards. - `optimize_living-artifacts-architect`: Step-by-step guidance for authoring `llms.txt`, `llms-full.txt`, and `ard.json`. - `optimize_agent-auth-handbook`: Blueprint for RFC 9728 machine-to-machine authentication contracts (`auth.md`). - `optimize_streamable-mcp-builder`: Architectural guide for deploying streamable HTTP MCP servers. - `optimize_flight-simulator-replay`: Diagnostic playbook for analyzing agent bottlenecks. ### Resources - `skill://glintbase/agent-readiness`: Master ARS 3.0 Playbook. - `skill://glintbase/living-artifacts`: Living Artifacts Specification. - `skill://glintbase/agent-auth`: Agent Authentication Handbook. - `glintbase://schemas/ars3-spec`: Complete ARS 3.0 evaluation schema. - `glintbase://standards/auth-contract`: RFC 9728 machine auth template. --- ## 🚀 Quick Start ### Option A: Smithery CLI ```bash npx -y @smithery/cli install @glintbase/mcp --client claude
Score: 100/100
Handshake verified by our own probe.
Tools · 5
Execute a full ARS 3.3 agent-readiness audit on a target URL. Probes Discovery, Access, Usability and Payments; checks that cannot be verified are reported as skipped and excluded from the score.
Fetch ultra-compact ARS 3.3 score card (<200 tokens). Ideal for quick checks and CI status monitoring without burning agent context.
Run the Glintbase Agent Flight Simulator across synthetic coding personas. Returns multi-modal output with a visual Journey Tree graphic, telemetry, and replay link.
Discover machine-readable entrypoints (robots.txt, llms.txt, ard.json, auth.md, OpenAPI, sitemap). Probes availability, syntax, and compliance.
Retrieve full markdown skill playbook for agent readiness (e.g. "agent-readiness", "living-artifacts-architect", "agent-auth-handbook", "mcp-server-hardening").
Resources · 11
Master framework for auditing, scoring, and optimizing web applications and APIs across the 4 ARS 3.3 layers.
skill://glintbase/agent-readiness
Guidelines and templates for designing high-density, token-budgeted llms.txt, llms-full.txt, and ARD manifests.
skill://glintbase/living-artifacts
Autonomous machine authentication standards, WorkOS auth.md schema, OAuth2 client credentials, and HTTP headers.
skill://glintbase/agent-auth
Production architecture for deploying streamable HTTP MCP routes in Next.js and Express under Anthropic MCP 2024-11-05.
skill://glintbase/streamable-mcp
Client-side agent tool registration via W3C draft window.modelContext and HTML form tool tags.
skill://glintbase/webmcp
Techniques for minimizing prompt bloat and eliminating ambiguous OpenAPI / MCP schemas that trigger hallucinations.
skill://glintbase/token-tax-schema
Interpreting synthetic coding agent runs across Claude Code, Cursor, Perplexity, and Swarm personas.
skill://glintbase/flight-simulator
Setting up automated GitHub Actions to block score regressions and post interactive PR comments.
skill://glintbase/ci-gate
Zero-Trust machine identity, soft-404 anti-hallucination barriers, mutation idempotency locks, and OWASP/ISO 42001 compliance standards.
skill://glintbase/agent-governance
ARS 3.3 check registry across 4 layers (Discovery, Access, Usability, Payments)
glintbase://schemas/ars3-spec
WorkOS machine-to-machine authentication template for auth.md
glintbase://standards/auth-contract
Prompts · 9
Master framework for auditing, scoring, and optimizing web applications and APIs across the 4 ARS 3.3 layers.
Guidelines and templates for designing high-density, token-budgeted llms.txt, llms-full.txt, and ARD manifests.
Autonomous machine authentication standards, WorkOS auth.md schema, OAuth2 client credentials, and HTTP headers.
Production architecture for deploying streamable HTTP MCP routes in Next.js and Express under Anthropic MCP 2024-11-05.
Client-side agent tool registration via W3C draft window.modelContext and HTML form tool tags.
Techniques for minimizing prompt bloat and eliminating ambiguous OpenAPI / MCP schemas that trigger hallucinations.
Interpreting synthetic coding agent runs across Claude Code, Cursor, Perplexity, and Swarm personas.
Setting up automated GitHub Actions to block score regressions and post interactive PR comments.
Zero-Trust machine identity, soft-404 anti-hallucination barriers, mutation idempotency locks, and OWASP/ISO 42001 compliance standards.
How to use
Claude Desktop or claude.ai: add a custom connector (Customize → Connectors; on Team and Enterprise plans an owner adds it under Organization settings → Connectors) with this server URL: https://glintbase.dev/api/mcp
Cursor or Cline: add to your MCP config:
{
"mcpServers": {
"glintbase": {
"url": "https://glintbase.dev/api/mcp",
"transport": "http"
}
}
}