Skip to content

Choosing a method

Pick the connector that matches your stack. Every method targets the same ingestion contract, so the only question is which one is least work for you.

Decision table

Your situationUse thisPage
Your site runs WordPress, Shopify, Drupal, Magento, Joomla, Ghost, Laravel, Symfony, Next.js, Nuxt, Wix, Webflow, or DjangoThe matching prebuilt connectorSee the rows below
You build in TypeScript or JavaScript and want a typed client plus a CLIThe Connector SDKConnector SDK
Your stack has no prebuilt connector and you want full control, in any languageThe generic API connector (raw HTTP)Generic API connector
You want an AI agent to do the integration for you, on any stackThe Universal AI PluginUniversal AI Plugin
You use Claude (Claude Code or the Agent SDK) and want one-command publishingThe Claude SkillClaude Skill
You want an AI client to read your published node, not write to itThe MCP serverMCP server

By platform

Match your platform to its prebuilt connector.

I runUse this
WordPress or WooCommerceWordPress plugin
ShopifyShopify connector
Drupal (with or without Commerce)Drupal module
Magento or Adobe CommerceMagento module
JoomlaJoomla plugin
GhostGhost connector
LaravelLaravel package
SymfonySymfony bundle
Next.jsNext.js package
NuxtNuxt module
DjangoDjango package
WixWix connector
WebflowWebflow connector

By situation

"I have a JavaScript or TypeScript app"

If you are on Next.js or Nuxt, use those connectors directly: they wire build-time adapters and server routes for you. For any other JS/TS app (a custom API, a static-site generator, a script), use the Connector SDK: it gives you a typed client, the signing primitives, and the bainquet-ingest CLI for cron jobs and CI.

"There is no connector for my stack"

You have two paths.

  • The generic API connector. Any language that can do HTTP, HMAC-SHA256, and HKDF can integrate directly against the contract. This is the lowest-level method and has no dependency. See Generic API connector.
  • The Universal AI Plugin. Hand an AI agent the scaffold (or its URL) and ask it to publish your project. The agent reads the recipe, discovers your content, maps it, signs requests, and streams them in. See Universal AI Plugin.

"I use Claude Code"

Install the Claude Skill (bainquet-pipeline). With the skill installed, ask Claude to publish, sync, or connect your project. It handles token use, content mapping, HMAC signing, batch posting, verification, and incremental sync.

"An AI client needs to read my node"

The methods above are the write side. The MCP server is the read side: it lets an MCP-capable AI client retrieve a published node's knowledge through tool calls, scoped to a website. Published node files are also plain static URLs on the CDN, so any consumer can fetch llms.txt, ai.json, and the JSONL files directly.

How the methods relate

The prebuilt connectors and the Universal AI Plugin are complementary: hand-built plugins for popular platforms, and the universal agent path for everything else. All write methods (the connectors, the SDK, the generic recipe, the Universal AI Plugin, and the Claude Skill) target the same contract and are checked against the same conformance reference. See Conformance.

Owner-controlled structured data for AI.