Fully Agentic AI Browser that can Act

Today’s top AI Highlights:

1. Opensource infrastructure to use, build, and scale MCP easily

2. Fully managed MCP-based RAG for your AI agents

3. Opensource version of OpenAI Advanced Voice Mode with ~500ms latency and real-time streaming

4. The first agentic browser that can think, browse, and act in shadow

& so much more!

Read time: 3 mins

AI Tutorial

Integrating travel services as a developer often means wrestling with a patchwork of inconsistent APIs. Each API—whether for maps, weather, bookings, or calendars—brings its own implementation challenges, authentication systems, and maintenance burdens. The travel industry's fragmented tech landscape creates unnecessary complexity that distracts from building great user experiences.

In this tutorial, we’ll build a multi-agent AI travel planner using MCP servers as universal connectors. By using MCP as a standardized layer, we can focus on creating intelligent agent behaviors rather than getting bogged down in API-specific quirks. Our application will orchestrate specialized AI agents that handle different aspects of travel planning while using external services through the MCP. We'll be using the Agno framework to create and orchestrate our team of specialized AI agents.

We share hands-on tutorials like this every week, designed to help you stay ahead in the world of AI. If you're serious about leveling up your AI skills and staying ahead of the curve, subscribe now and be the first to access our latest tutorials.

Don’t forget to share this newsletter on your social channels and tag Unwind AI (X, LinkedIn, Threads, Facebook) to support us!

Latest Developments

Build and Use Production-Ready MCP Servers at Scale 🧰 🖇️ 🖥️

Klavis AI is an open-source platform that simplifies integration with Model Context Protocol (MCP) servers and clients. It provides hosted infrastructure and APIs to spin up production-ready MCP servers instantly.

It solves common problems devs face with MCPs—like missing auth, lack of hosted options, and needing to write custom client code. With Klavis, you can spin up an MCP server in seconds using an API, and connect it to your AI app without building glue code. It also ships with open-source clients for web, Slack, and Discord to test and deploy quickly.

Key Highlights:

1. Instant MCP Integration - Deploy production-ready MCP servers in under a minute through simple API calls. No wrestling with complex setups or reliability issues when connecting AI applications to external tools. Just one API call creates your server instance with a dedicated URL that's ready to use immediately.

2. Built-in Authentication - Klavis handles OAuth and multi-tenant authentication out of the box, securing connections between your AI applications and services without you having to manage tokens or write custom auth code.

3. Multiple Clients - Access your MCP servers through ready-made clients for Web, Slack, and Discord. These open-source interfaces let users interact directly with AI applications powered by MCP services, making it easier to build user-facing AI tools that work where your users already spend their time.

4. Deployment Options - Choose between fully hosted services with API access or run everything yourself using the open-source codebase. The hosted option gives you instant scalability with 100% connection guarantees on dedicated infrastructure, while the self-hosted route offers maximum control and customization for specialized needs.

Deploy RAG with MCP Support in Minutes 📚🔍🤖

CustomGPT now offers a hosted MCP Server that connects directly to its production-grade RAG stack—no infra setup required. Developers can deploy an MCP-compliant endpoint in under 2 minutes and plug it into tools like Claude Desktop, Cursor, or Zapier.

You connect your data (Google Drive, PDFs, Notion, etc.), enable the MCP server, and immediately get an endpoint compatible with tools like Claude Desktop, Cursor, n8n, and LangChain. It’s included in all CustomGPT plans, even the free trial, and is built on their top-ranked RAG stack for business-doc accuracy.

Key Highlights:

1. Deploy in 1-2 minutes - You can deploy a fully managed RAG system via a standard MCP endpoint in under 2 minutes. It handles ingestion, indexing, embeddings, reranking, and scaling—no Kubernetes, no TLS renewal, no vector DB config needed.

2. Compatible with MCP clients - Works with any MCP client including Claude Desktop, ChatGPT (with plugin), Cursor IDE, workflow tools like n8n and Zapier, and agent frameworks like LangGraph and AutoGen. Connect once and use everywhere without building custom integrations.

3. Real-time sync - Supports Google Drive, Notion, Confluence, PDFs, and more. Indexing is near-instant, and the system auto-refreshes whenever content changes. It’s SOC-2 Type 2 certified, with encrypted access via bearer tokens and secure SSE streams.

4. Full feature access during trial - You can test the entire setup—RAG quality, agent responses, and end-to-end data flow—using up to 1,000 documents on the free trial. No feature gating, no hidden infra costs, and works with your existing CustomGPT agents.

Quick Bites

Amazon’s AI coding agent, Amazon Q Developer, is now available in VS Code with a new agentic coding experience. It can read and write local files, run terminal commands, and help you build, debug, or refactor through natural conversation. The assistant understands your full codebase context and gives you the option to apply code changes automatically or step-by-step. Available at no extra cost for both Free and Pro tier users.

Refact.ai’s opensource agent just hit 59.7% on SWE-bench Lite, topping the leaderboard by solving 179 out of 300 real GitHub tasks—fully autonomously. No manual steps, no scripted workflows—just Claude 3.7 Sonnet orchestrating the logic, o4-mini handling reasoning, and a dynamic toolset for repo exploration, editing, and test validation.

Developers can try it today in VS Code or JetBrains—Refact.ai is built to plug straight into your workflow and get actual software tasks done end-to-end.

We came across this interesting opensource real-time AI voice chat project. It lets you have natural voice conversations with an AI in your browser, with replies coming back in about 500ms. It captures your voice, transcribes it using RealtimeSTT, runs it through an LLM like Ollama or OpenAI, and streams back a spoken response using RealtimeTTS—all without waiting. You can even interrupt mid-sentence, just like a real conversation. Built with FastAPI, WebSockets, and Vanilla JS, it’s a full-stack voice interface setup ready to be built upon.

Tools of the Trade

1. Fellou: The first agentic browser that can think, browse, and act across public and private web pages using your own device and login. It runs in a shadow window, plans and executes multi-step tasks like emailing, scheduling, or reporting, and supports deep search, all without disrupting your main screen.

2. Peek: AI "Spotify wrapped" for your money. Personal finance app where an AI agent tracks your spending, sends weekly digests, and nudges you toward better money habits. It replaces charts and dashboards with conversational check-ins, goal-based insights, and automated tracking.

3. Ridvay Code: AI developer agent for VS Code that helps with code generation, debugging, refactoring, and documentation for all languages, directly inside your editor. It supports natural language prompts to generate or explain code, restructure large codebases, suggest bug fixes, and create automated tests.

4. Awesome LLM Apps: Build awesome LLM apps with RAG, AI agents, MCP, and more to interact with data sources like GitHub, Gmail, PDFs, and YouTube videos, and automate complex work.

Hot Takes

1. real software engineers are unemployed ~
   Kevin Naughton Jr.
   
   

2. OpenAI has to level up to Sonnet 3.7 in coding
   Anthropic has to level up to o3 in planning
   Gemini has to get better at instruction following
   Open source is to get better at everything!
   A lot of room for LLMs to still grow ~
   Bindu Reddy

That’s all for today! See you tomorrow with more such AI-filled content.

Don’t forget to share this newsletter on your social channels and tag Unwind AI to support us!

Unwind AI - X | LinkedIn | Threads | Facebook

Awesome LLM Apps | Sponsor Us

PS: We curate this AI newsletter every day for FREE, your support is what keeps us going. If you find value in what you read, share it with at least one, two (or 20) of your friends 😉