I normally like to open the newsletter with general news in the world of AI, but I've been heads-down with the rest of the Langflow team working on the newly released Langflow version 1.6. With features like OAuth for MCP, a Docling powered file component, and compatibility with the OpenAI API, it's been worth it.

What else has been going on then? OpenAI just hosted their DevDay event with a pile of new releases, while Anthropic had people queue for 2 hours for free hats. Everyone else seems to be launching AI search APIs. Thanks for checking it all out in this edition of AI++!

Phil Nash
Developer relations engineer for Langflow

🛠️ Building with AI, Agents & MCP

​Langfow 1.6 launched​

This is a significant new release as it bundles Docling to make for advanced file parsing. It also brings MCP Composer into the mix to power OAuth protection for your Langflow-powered MCP servers, integrated Traceloop for another observability option, tidied up the UI around components and MCP servers, and added a bunch of new components. Go try Langflow 1.6 now!

Anthropic have been busy

Not only has Claude Sonnet 4.5 been released, but the Claude developer platform released tools for memory and context editing, and there's now a Claude Agent SDK available in TypeScript and Python.

OpenAI have been busy too

OpenAI and Stripe announced Agentic Commerce which is a protocol for commerce between buyers, AI agents, and businesses. It is powering Buy It In ChatGPT and you can read more about how Stripe and OpenAI worked together to create this protocol.

Then OpenAI DevDay happened and they announced Apps in ChatGPT, AgentKit, new models in the API including GPT-5 Pro, gpt-realtime-mini, and gpt-image-1-mini. Even Codex got an API too. All the announcements are on the DevDay site.

AI web search APIs abound

First Ollama announced that they were making a web search API available, then a day later Perplexity announced their search API. These join Tavily, Brave, and Serper as good options to ground your agents with real search results.

​Code Mode: the better way to use MCP​

Cloudflare have taken a different approach to MCP, arguing that models have seen far more TypeScript in the world than training data on calling MCP tools. So they turned MCP tools into TypeScript functions and found that agents could handle more tools, more complex tools, and string tool calls together in custom code without having to go to the model with each individual result. You can try this out in the Cloudflare Agents SDK now.

​Exploring Active Agent, or can we build AI features the Rails way?​

I know most AI news is about Python or TypeScript, but I have a soft spot for Ruby as a language, so I was delighted to see that there are people out there trying to build agents in the Rails way.

🗞️ Other news

• OpenAI also introduced ChatGPT Pulse. It's not something you can use as a developer, but it does change the story around user interactions with AI agents when they can start the conversation.
• Over at MongoDB they've been thinking about why you need to engineer the memory of multi-agent systems.
• There are rumors that Apple are building MCP support into iOS, iPadOS and macOS.
• Ever wondered which table format LLMs understand the best? This battle between 11 formats from CSV to XML is worth a read.
• On the stranger end of the AI news this week is that Anthropic didn't have to try very hard to convince people to queue for 2 hours for a free coffee and hat. I do want my own "thinking cap" though...
• And finally an artist painted a banana and then used ChatGPT to bring it to life. Check out the banana portal, something that simply would not have existed without AI.

🧑‍💻 Code & Libraries

• ​Airweave is an open source tool that can help agents search within any application.
• Google released an API for Jules, so now you can build agents that can build agents.
• Also from Google, the Data Commons MCP server unlocks huge amounts of public data for your agents.

🔦 Langflow Spotlight

I've got to bring it back to the File component this week, as it just got new superpowers in Langflow 1.6. Add a file to the component, flip the Advanced Parser switch, then pick your pipeline and OCR engine and you now have the power of Docling waiting to accurately parse your documents for your agents.

🗓️ Events

It's a busy couple of weeks for the Langflow team, we're going to be all over the place and it would be great to meet you at any of these events!

October 9th, 5:00 PM PDT, San Francisco - It's SF Tech Week and you can catch Carter from the Langflow team, along with OpenAI, Vercel and Stytch at the MCP & AI Agents: Drinks + Panel.

October 14th, 6:30 PM BST, London - Come on down to JSMonthly where I'll be talking about Web AI.

October 14th, 11:00 AM PDT – Security is the topic of The Flow as David and Carter chat with Max Gerber from Stytch. Catch it live or subscribe to the podcast to get the recording.

October 15th–16th, London – join the Langflow team at AI for the rest of us. Promising two days of inspiring and practical sessions that demystify jargon, share real world stories of AI in action, and include developer focused sessions and hands-on workshops, this will be a fantastic community event. You can save 20% off the cost of a ticket with the code LANGFLOWCREW. And look out for me in the keynote session on 16th building an agent in 20 minutes.

October 15th-16th - If you're in Munich, then check out the InfoQ Dev Summit where Tejas will be giving the keynote on the 16th on AI Innovation in 2025 and beyond.