@agentreadyweb/mcp

What is AgentReady?

AgentReady is a hosted capability layer that makes any website discoverable and usable by agents through MCP.

Paste a URL → AgentReady crawls the site, generates a spec-compliant llms.txt, and hosts a live /ask RAG endpoint and MCP server. Any MCP-compatible agent client can then discover the site, ask natural-language questions, inspect capabilities, and create grounded read-only plans with citations.

The problem it solves: AI agents using web_fetch fetch one page at a time, get empty HTML from JavaScript SPAs (React, Next.js, Vue), and hallucinate when the answer spans multiple pages. AgentReady indexes the whole site, handles JS rendering, and retrieves across pages — so agents get the right answer instead of a confident wrong one.

What's already indexed: Browse the live AgentReady directory → or index any public site yourself.

Key properties:

• Works on any public URL — static sites, React/Next.js SPAs, Docusaurus, GitBook, custom engines
• No account required to index your first site
• Shared index — one team member submits a site, everyone on the team can query it instantly
• Handles JS-rendered pages that web_fetch returns empty for
• Capability manifests expose freshness, schemas, endpoints, and read-only limits
• Grounded plans return evidence, risks, confirmation requirements, and durable receipts

Built with Codex and GPT-5.6

Codex, powered by GPT-5.6, was used throughout development as an engineering copilot. It helped extend and test the MCP server, automatic indexing flows, capability manifests, safe action planning, CLI commands, documentation, and production integrations.

The MCP package is model-agnostic at the client boundary. GPT-5.6 supported development, testing, and iteration without being required by the end user's MCP client.

CLI

The same package doubles as a CLI — no install, no account:

# Agent-readiness report card (llms.txt, sitemap, robots, JS-rendering, index status)
npx @agentreadyweb/mcp grade yourdocs.com

# Ask any site a question, get a cited answer (auto-indexes new sites in ~60s)
npx @agentreadyweb/mcp ask stripe.com "what is the test card number?"

# Index or re-crawl a site
npx @agentreadyweb/mcp index yourdocs.com
npx @agentreadyweb/mcp refresh yourdocs.com

For CI, dashboards, or scripts, add --json to receive the raw report on stdout (progress remains on stderr):

npx @agentreadyweb/mcp grade yourdocs.com --json | jq '.grade, .score'

The command still exits 1 when the grade is below B.

grade exits non-zero below a B, so you can use it as a CI gate. refresh in your docs deploy pipeline keeps the index fresh automatically:

# GitHub Actions — after your docs deploy step
- run: npx @agentreadyweb/mcp refresh yourdocs.com

---

Connect any MCP-compatible client to AgentReady:

Local MCP bridge (stdio)

For clients that use an mcpServers configuration, add the AgentReady bridge:

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

Restart your client. You'll have seven tools available:

• list_sites — see all indexed websites
• get_site_capabilities — index on demand, then inspect a site manifest, freshness, schemas, and endpoints
• ask_site — query any site with cited, multi-page answers
• plan_site_action — index on demand, then create a grounded, read-only plan and receipt
• submit_site — index any website so it can be queried
• refresh_site — re-crawl a site, or perform its initial index when it is new
• rate_answer — submit quality feedback

Cursor

Add to ~/.cursor/mcp.json:

{
  "agentready": {
    "command": "npx",
    "args": ["-y", "@agentreadyweb/mcp"]
  }
}

Project or terminal setup

npx @agentreadyweb/mcp

Or add the same agentready server to your client's project configuration to share it with your team:

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

VS Code (GitHub Copilot agent mode)

Requires VS Code 1.99+ with the GitHub Copilot extension. Create .vscode/mcp.json in your project root:

{
  "servers": {
    "agentready": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@agentreadyweb/mcp"]
    }
  }
}

Copilot's MCP tools are only available in agent mode. Commit this file to share with your team.

Windsurf / Zed / other clients

Any MCP client that supports stdio transport works the same way — use npx -y @agentreadyweb/mcp as the command.

WebMCP (no install)

If your client supports HTTP transport, connect directly by URL — no npm required:

https://www.agentready.it.com/api/mcp

Docusaurus plugin

If your docs site uses Docusaurus, auto-index on every build:

npm install @agentreadyweb/docusaurus-plugin

// docusaurus.config.js
plugins: [
  ['@agentreadyweb/docusaurus-plugin', { domain: 'docs.yoursite.com' }]
]

MkDocs plugin

If your docs site uses MkDocs, auto-index on every build:

pip install mkdocs-agentready

# mkdocs.yml
plugins:
  - search
  - agentready

If site_url is not set, specify the domain explicitly:

plugins:
  - agentready:
      domain: docs.yoursite.com

Starlight plugin

If your docs site uses Starlight (Astro), auto-index on every build:

npm install starlight-agentready

// astro.config.mjs
import agentready from 'starlight-agentready'

export default defineConfig({
  site: 'https://docs.yoursite.com',
  integrations: [
    starlight({
      plugins: [agentready()],
    }),
  ],
})

Sphinx extension

If your docs site uses Sphinx, auto-index on every build:

pip install sphinx-agentready

# conf.py
extensions = [
    "sphinx_agentready.extension",
]

# Domain is inferred from html_baseurl automatically, or set explicitly:
agentready_domain = "docs.yoursite.com"

Available tools

list_sites

Lists all websites currently indexed by AgentReady with their titles and page counts. Use this to check if a domain is already available before submitting it.

get_site_capabilities

get_site_capabilities(domain: string)

Returns the site's capability manifest, freshness state, schemas, and available HTTP/MCP endpoints.

plan_site_action

plan_site_action(domain: string, request: string)

Creates a grounded, read-only plan with evidence, risks, confirmation requirements, and a durable receipt. It does not execute side effects.

submit_site

submit_site(url: string)

Index any website with AgentReady. Takes ~60 seconds. Handles static sites, server-rendered pages, and JavaScript-heavy SPAs via a four-layer pipeline. Once done, query it with ask_site.

Example: submit_site("https://docs.example.com")

ask_site

ask_site(domain: string, query: string, url?: string)

Ask a question about any website and get a cited answer grounded in its content. Synthesises information across multiple pages. If the site isn't indexed yet, AgentReady crawls and indexes it automatically before answering (~60s).

Example: ask_site("stripe.com", "What are the fees for card payments?")

refresh_site

refresh_site(domain: string)

Force a full re-crawl of a site to pick up new or changed content. If the site is new, AgentReady performs its initial index automatically. Takes ~60 seconds.

Example: refresh_site("docs.example.com")

rate_answer

rate_answer(domain: string, rating: number, request_id?: string, comment?: string)

Submits 1–5 quality feedback, optionally tied to the exact ask_site request.

Deploy webhook

Automatically re-index your docs on every deploy. No auth required — rate limited to once per hour per domain.

curl -X POST https://www.agentready.it.com/api/webhook/refresh \
  -H "Content-Type: application/json" \
  -d '{"domain": "docs.yoursite.com"}'

Or pass the domain as a query param (works with Vercel/Netlify form-encoded webhook payloads):

https://www.agentready.it.com/api/webhook/refresh?domain=docs.yoursite.com

How indexing works

AgentReady uses a layered approach to handle any public website:

1. llms.txt / llms-full.txt — if the site publishes one, it's used as a high-quality structured content source
2. Standard HTML crawl — fetches up to 10 pages via sitemap or link crawling, extracts clean text
3. __NEXT_DATA__ extraction — for Next.js apps, parses server-side rendered data embedded in the HTML
4. Jina Reader fallback — for JS-only SPAs that return empty HTML, uses a remote rendering service to extract content

Sites behind authentication or with no public HTML content cannot be indexed.

Environment variables

Variable	Default	Description
AGENTREADY_MCP_URL	https://www.agentready.it.com/api/mcp	Override the MCP endpoint (for self-hosted)