Opensource Gemini CLI Agent with 1M Context

Today’s top AI Highlights:

1. Google’s opensource CLI agent with ridiculously free limits

2. AI agents that scale with a zero-cost abstraction

3. Opensource Perplexity clone to get answers with citations and live data 

4. Claude just got its own app store

5. First opensource AI Design Agent that lives inside your IDE

& so much more!

Read time: 3 mins

AI Tutorial

We've been stuck in text-based AI interfaces for too long. Sure, they work, but they're not the most natural way humans communicate. Now, with OpenAI's new Agents SDK and their recent text-to-speech models, we can build voice applications without drowning in complexity or code.

In this tutorial, we'll build a Multi-agent Voice RAG system that speaks its answers aloud. We'll create a multi-agent workflow where specialized AI agents handle different parts of the process - one agent focuses on processing documentation content, another optimizes responses for natural speech, and OpenAI's text-to-speech model delivers the answer in a human-like voice.

Our RAG app uses OpenAI Agents SDK to create and orchestrate these agents that handle different stages of the workflow.

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) to support us!

Latest Developments

Google Brings Gemini 2.5 Pro Agent to CLI 🧑‍💻🤖🦾

AI agents have officially escaped the IDE prison. The holy trinity of AI - OpenAI, Anthropic, and now Google has their AI coding agents living in the Terminal.

Google has released Gemini CLI, an open-source AI agent that brings Gemini 2.5 Pro and its massive 1 million token context window directly to your Terminal. This agent connects to your tools, understands your code, and helps you build applications and automate operational tasks, like querying pull requests.

Google's betting on radical accessibility: industry-leading free usage limits, native Windows support without WSL, and Apache 2.0 licensing that lets developers actually own their tools.

Key Highlights:

1. Free Access - Gemini CLI offers 60 requests per minute and 1,000 requests daily at zero cost with a free Gemini Code Assist license, crushing the competition with the largest usage allowance for individual developers.

2. Context Management - Hierarchical memory system learns from GEMINI.md files while git-aware context automatically respects .gitignore, eliminating the manual file selection headaches plaguing other CLI agents.

3. Multi-Modal - Processes images, PDFs, and live web search results through Google Search grounding, giving it real-time external context that static models can't match for current information needs.

4. Extensibility - Native MCP server support, shareable custom extensions, and observability with traces and metrics make it the most enterprise-ready CLI agent despite being completely free.

Set up within a minute with npm install -g @google/gemini-cli.

AI Agents as Pure Functions With Automatic Checkpointing 📋🔄

Your AI agents crash at the worst possible moments, don't they? Right when they're halfway through a complex multi-step task, or worse, when they're waiting for human approval that might take hours.

Usually, AI agent frameworks treat this as an acceptable trade-off, but it doesn't have to be.

Pickaxe is a TypeScript library that brings durable execution to AI agents without the framework baggage. It automatically checkpoints your agent's state so they can recover from failures, wait for external events indefinitely, and scale to handle millions of tasks. The beauty lies in its simplicity - agents are just functions you write, not framework constructs you inherit.

Key Highlights:

1. Bulletproof Execution - Agents automatically checkpoint their progress, so when your infrastructure fails mid-execution, they resume exactly where they left off without losing work or state.

2. No Framework Lock-in - Your agents are plain TypeScript functions that integrate with existing codebases, giving you full control over LLM calls, memory systems, and business logic.

3. Infinite Scale - Runs billions of tasks monthly with built-in support for massive parallelism, distributed execution, and intelligent scheduling across machine fleets.

4. Production-First - Ships with rate limiting, retry mechanisms, concurrency control, and works on any container platform from Railway to Kubernetes.

Quick Bites

Looking for a Perplexity alternative that you can further build upon? Firecrawl just released Fireplexity, an opensource AI search engine that delivers answers with real-time web citations using their scraping API, powered by GPT-4o-mini. Comes with smart follow-up questions, automatic stock charts via TradingView integration, and streaming responses.

Cloudflare has released Code Sandboxes, the easiest way to run AI-generated code on demand. It lets you spin up isolated containers where your Workers can safely execute untrusted LLM code without compromising your infrastructure. The API is simple - just a few lines of code to get started, with built-in methods for file operations, git integration, and command execution.

Claude just got its own app store, but you're the developer. Anthropic rolled out a dedicated home for building, organizing, and sharing your AI creations. You can now embed Claude's API directly into any artifact, creating fully functional apps powered by AI that respond to user input.

When you share these AI-powered artifacts, viewers authenticate with their own Claude accounts, keeping usage off your subscription.

Tools of the Trade

1. SuperDesign.dev: Opensource Design Agent that integrates directly into IDEs like Cursor, Windsurf, and VS Code to generate UI mockups, components, and wireframes from simple prompts. It runs multiple design agents in parallel, so you can explore 10 different design variations simultaneously and iterate on them.

2. Scenario: AI agent testing framework for end-to-end testing of AI agents by simulating realistic user interactions across different scenarios and edge cases. Provides three levels of testing - unit tests, component evals, and end-to-end agent simulations.

3. Chord: Collaborative chat platform that allows multiple users to work together with various AI models (OpenAI, Anthropic, Gemini, etc.) in shared conversation threads with persistent context.

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. why tf does a gpt wrapper need millions in funding? your entire tech stack is three API calls and cursor writes all your code ~
   Amogh Chaturvedi
   
   

2. There will be no edge in talent in AI

   There will be no edge in compute in AI

   There will be no edge in models in AI

   Will be

   Data

   Distribution

   Integration

   Google et al will drive generalised AI to zero marginal cost & we will make open variants of cutting edge open & available ~
   Emad Mostaque

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

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 😉