Building a Telnyx Adapter for Chat SDK

Jan Herklotz — Fri, 13 Nov 2026 00:54:47 +0000

I built an open-source adapter that connects Telnyx to Chat SDK, so you can build bots that send and receive SMS and MMS messages.

The adapter supports group MMS, scheduled messages, media attachments, and delivery receipts. It also verifies incoming Telnyx webhooks using Ed25519 signatures.

Getting started

Install the adapter alongside Chat SDK and a state adapter:

npm install chat-adapter-telnyx chat @chat-adapter/state-memory

Then configure it with your Telnyx API key, phone number, and public key:

import { createMemoryState } from "@chat-adapter/state-memory";
import { Chat } from "chat";
import { createTelnyxAdapter } from "chat-adapter-telnyx";

export const bot = new Chat({
  userName: "mybot",
  adapters: {
    telnyx: createTelnyxAdapter({
      apiKey: process.env.TELNYX_API_KEY!,
      phoneNumber: process.env.TELNYX_PHONE_NUMBER!,
      publicKey: process.env.TELNYX_PUBLIC_KEY!,
    }),
  },
  state: createMemoryState(),
});

bot.onDirectMessage(async (thread, message) => {
  await thread.post(`You said: ${message.text}`);
});

You’ll also need to route incoming Telnyx webhooks to bot.webhooks.telnyx. The README includes a Next.js example and setup details.

Why I built it

Chat SDK provides a common interface for building chat bots across platforms. I wanted to use that same approach for SMS and MMS through Telnyx.

The adapter is open source and available on GitHub and npm. Feedback, issues, and contributions are welcome.

Building MCP Servers for Product Search and Availability

Jan Herklotz — Fri, 13 Nov 2026 00:54:04 +0000

I recently built a few MCP servers that let AI assistants search listings, find products, and check prices and availability.

The idea was simple: instead of opening several websites and searching each one manually, I wanted to ask an assistant to find what I was looking for.

So far, I’ve built servers for:

Apple Store: products, configurations, prices, delivery estimates, and pickup availability
MediaMarkt, MediaWorld, and Saturn: product search, store search, and pickup availability
MaxGaming: gaming products, prices, and availability across storefronts
Kleinanzeigen: searching and reading listings

What I learned

MCP gives each server a consistent way to expose tools to an assistant. The more interesting work is figuring out how to get useful data from each site and return it in a form the assistant can work with.

Every site is different. Product pages expose options and variants differently, availability depends on the storefront or location, and search results don’t always match the query as closely as you’d expect.

There’s also a difference between “unavailable” and “I couldn’t check.” If a request fails or data is missing, the assistant needs to know rather than treating it as an out-of-stock result.

What’s next

I’m interested in making these servers easier to use and improving coverage where product pages or storefronts behave differently.