AI++ // multi-agent systems are go


Privacy has been a hot topic for consumer AI applications this month. Users have been finding out that chats with OpenAI, Grok, and MetaAI that they shared, were being indexed by Google and showing up in search results. Sharing a chat with your friend doesn't mean you want to share it with every search engine out there, so make sure your users know what they're doing and don't forget your robots.txt.

Multi-agent systems are the theme for this newsletter, with tutorials on how to build them and many, many lessons learned from having built them.

Phil Nash
Developer relations engineer for Langflow

πŸ› οΈ Building with AI, Agents & MCP

​How to Build a Deep-Research Multi‑Agent System​

Get five agents doing all the business of planning, sourcing, summarizing, reviewing, and reporting for you and explore why multi-agent systems might be right for your project.

​Teaching GPT-5 to Use a Computer​

One developer's project to build an agent that controls your computer in front of you using a hierarchical agent architecture, not bad for a hackathon project! The post goes into great depth into how it works and how it was optimized and trained.

​Best Practices for Building Agentic AI Systems: What Actually Works in Production​

Are stateless subagents in a two-tier system the secret for a powerful agent application? After much experimentation, this developer thinks so.

​What makes Claude Code so damn good (and how to recreate that magic in your agent)!?​

This article breaks down all the neat tricks Claude Code uses to create a delightful AI agent experience. Turns out it also only uses one level of subagent, so maybe there's something in that.

​APIs don't make good MCP tools​

APIs and MCP servers have two very different consumers, so you shouldn't just wrap one in the other. This article shows why and includes tips for your MCP server, check out why CSVs are back in fashion.

​Wicked Python trickery - dynamically patch a Python function's source code at runtime​

Did you know you can dynamically change a Python function’s source code at runtime? What if you could use an LLM to write new code and then dynamically execute it within the context of your application? It might not be a secure idea, but it’s definitely possible!

πŸ—žοΈ Other news

​MCP Vulnerabilities Every Developer Should Know​

OK, I think I've put an article about MCP vulnerabilities into every edition of AI++ so far, but it might be the most important thing you can consider when building your own MCP server, so I'm probably going to continue.

​Context Rot​

This is a great bit of research on how LLM performance can drop off as the context increases. It's quite in-depth, so if you want the quick version, the video summary is worth a watch.

πŸ§‘β€πŸ’» Code & Libraries

πŸ”¦ Langflow Spotlight

The spotlight is normally a place to show off one of the great components you can use in your flows, but Langflow recently crossed 100,000 stars on GitHub so this week we're just celebrating instead πŸŽ‰

Thanks to all the users, contributors and maintainers. The mission to democratize AI goes on!

πŸ—“οΈ Events

11am PDT, August 27th – James Le from TwelveLabs joins David and Carter live on The Flow to investigate searching, analyzing, indexing and editing videos with TwelveLabs AI.

Can't make the live stream? Subscribe to The Flow wherever you get your podcasts.

September 18th–19th – Combine web, AI and community at CascadiaJS 2025 in Seattle. The Langflow team will be there, and you can get tickets for a magical 50% off with the promo code LANGFLOW_50.

September 20th – The AI party keeps rolling with the Cascadia AI Hack Day in Seattle. Hang out for a day to see what you can build using AI, Agents and MCP.

Enjoy this newsletter? Forward it to a friend.

2755 Augustine Dr, 8th Floor, Santa Clara, CA 95054
​Unsubscribe Β· Preferences​

AI++ newsletter

Subscribe for all the latest news for developers on AI, Agents and MCP curated by the Langflow team.

Read more from AI++ newsletter

The last couple of weeks has seen students booing commencement speakers at graduation ceremonies in Florida and Arizona when they mentioned AI. This is a visceral reaction to what they see as a threat to their careers. Meanwhile, developers working with AI are burning out and getting β€œBrain Fry” from doing more work at higher intensities without the same fulfillment. I wrote a bit about this myself, sharing that I found it hard to be proud of a useful little app that I built. AI is changing...

Working with LLMs is weird, but I never thought it would be as weird as OpenAI having to specifically tell their models not to talk about goblins, gremlins, raccoons, trolls, ogres, or pigeons. It raises so many questions. Thankfully after someone spotted the instructions in the Codex base instructions, OpenAI did give an explanation as to where the goblins came from. They never mentioned why raccoons and pigeons got caught up in the fantasy creature fascination though. In this edition of...

Is a token crunch coming? This week GitHub paused sign-us for GitHub Copilot Pro, Pro+ and Student plans, tightened up their usage limits, and removed Opus from their Pro plans. And today, Anthropic seemed to remove Claude Code from new Pro plans, though that has been reversed quickly. In general, while this is only seeming to affect individual plans related to coding agents, it could point to an inflection point where AI companies start considering how their pricing matches up to their...