AI2Web
An MCP server for Claude and ChatGPT that discovers AI-ready websites and acts on them. It fronts the AI2Web Discovery Network with three tools: find_sites, describe_site, and call_site_action. Actions are approval-gated and SSRF-guarded, and no site needs to adopt anything, Claude and ChatGPT already allow third-party MCP connectors. Install: npx @ai2web/connector.
README
<div align="center"> <a href="https://ai2web.dev"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/ai2web-foundation/.github/main/profile/ai2web-logo-white.svg"> <img alt="AI2Web" src="https://raw.githubusercontent.com/ai2web-foundation/.github/main/profile/ai2web-logo-black.svg" width="200"> </picture> </a> </div>
AI2Web JavaScript/TypeScript framework (@ai2web/*)
The reference framework for the AI2Web protocol. Describe your website's capabilities once; generate the manifest, routes, and adapters.
TypeScript, JavaScript, or React. The packages are authored in TypeScript but published as standard ESM JavaScript with bundled type declarations, so they run unchanged in plain JS. @ai2web/react adds hooks and an embeddable badge for React apps, and a framework-free <ai2w-badge> web component works in any HTML page with no build step. See examples/.
Packages
| Package | Status | Purpose |
|---|---|---|
@ai2web/core |
✅ scaffolded | Types, capability model, fluent manifest builder, validation + AI Readiness scoring, discovery client, export projections (toLlmsTxt / toAgentJson, RFC-0015), and the shared executeOperation guard (RFC-0006 §3: approval by risk tier, same-origin credentials, SSRF) that every adapter routes through. |
@ai2web/validator |
✅ scaffolded | ai2web validate <url> CLI → per-capability report + AI Readiness Score + tier. |
@ai2web/server |
✅ scaffolded | Framework-agnostic /ai2w, /ai2w/*, negotiation and action route handler, plus multi-surface projections (/llms.txt, /.well-known/agent.json, RFC-0015); Cloudflare Workers and Node adapters. |
@ai2web/mcp-bridge |
✅ scaffolded | Expose declared capabilities as an MCP server (usable in Claude/ChatGPT today). |
@ai2web/graphql-adapter |
✅ scaffolded | Project the declared model as a GraphQL schema (SDL + resolvers); GET actions become Query fields, writes become Mutations. |
@ai2web/acp-adapter |
✅ scaffolded | Drive a site's advertised ACP checkout transport (RFC-0005 Profile 1); the payment step stays approval-gated. |
@ai2web/ap2-adapter |
✅ scaffolded | Drive a site's advertised AP2 (Agent Payments Protocol) merchant transport: create a signed Cart Mandate, settle a Payment Mandate; the payment step stays approval-gated. |
@ai2web/nlweb-adapter |
✅ scaffolded | Query a site's advertised NLWeb (nlweb.ai) ask endpoint in natural language; routes through the shared guarded, SSRF-checked executor. |
@ai2web/openapi-adapter |
✅ scaffolded | Describe the declared actions as an OpenAPI 3.1 document; auth becomes security, risk/approval become x-ai2w-* extensions. |
@ai2web/connector |
✅ scaffolded | Agent-side connector: one MCP server that fronts the Discovery Network and acts on sites. |
@ai2web/react |
✅ scaffolded | React bindings: useDiscover / useValidate / useNegotiate hooks and an embeddable <Ai2wBadge>. React is an optional peer dependency. |
@ai2web/hono |
✅ scaffolded | Hono middleware: app.use("*", ai2w({ manifest, actions })) serves every AI2Web route and passes the rest through. Edge-native (Workers/Bun/Deno). |
@ai2web/astro |
✅ scaffolded | Astro integration (auto-wires the routes) + a createAi2wRoute endpoint factory. Serves from your manifest under SSR/hybrid output. |
@ai2web/nuxt |
✅ scaffolded | Nuxt module: add "@ai2web/nuxt" and point it at your config module; it registers the Nitro server routes for every AI2Web path. |
The executing adapters (MCP, GraphQL, ACP) derive their surface from the same manifest and run through the one executeOperation in @ai2web/core, so none can expose an undeclared capability or drift on the security contract; the OpenAPI adapter is descriptive but likewise describes only declared actions and preserves their auth/risk semantics. Conformance is enforced in CI by scripts/adapter-conformance.ts (MCP), scripts/adapter-graphql-acp.ts (GraphQL + ACP), and scripts/adapter-openapi.ts (OpenAPI).
Try the validator now (zero-dependency)
Before the TypeScript build, a runnable mirror lives at scripts/validate.mjs:
node scripts/validate.mjs ../ai2web-spec/examples/ecommerce.json # → AI Readiness Score 100/100
node scripts/validate.mjs https://some-site.com # fetches /ai2w with well-known fallback
Describe a site once (builder)
import { ai2web } from "@ai2web/core";
const manifest = ai2web({ name: "Example Store", url: "https://example.com", type: "ecommerce" })
.capability("content")
.capability("commerce", { endpoint: "/ai2w/products", checkout: true })
.transports({ mcp: { enabled: true, endpoint: "/ai2w/mcp" }, rest: { enabled: true, base: "/ai2w" } })
.auth({ methods: ["none", "oauth2"], oauth2: { pkce: true, scopes: ["read_products", "checkout"] } })
.consent({ requires_user_approval_for: ["purchase", "payment"] })
.events({ endpoint: "/ai2w/events", types: ["order.shipped", "price.drop"] })
.build();
Build (once dependencies are installed)
npm install
npm run build # tsc --build across workspaces
Licence
MIT.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.