Connect AI Agents to 100,000+ APIs

Today’s top AI Highlights:

1. Postman's new AI Agent Builder connects your Agents to 100,000+ APIs

2. Full stack agent infrastructure to build, deploy, and scale AI agents

3. $6,000 to run full DeepSeek R1 LLM with reasoning at home

4. New VC fund to back AI Agent startups with small-check strategy

5. A highly opinionated AI SDK with auto-model selection & built-in reasoning

& so much more!

Read time: 3 mins

AI Tutorials

Sales teams spend countless hours manually searching for and qualifying potential leads. This repetitive task not only consumes time but also results in inconsistent lead quality. Let’s automate this process to help sales teams focus on what matters most - building relationships and closing deals.

In this tutorial, we'll build an AI Lead Generation Agent that automatically discovers and qualifies potential leads from Quora. Using Firecrawl for intelligent web scraping, Phidata for agent orchestration, and Composio for Google Sheets integration, you'll create a system that can continuously generate and organize qualified leads with minimal human intervention.

Our lead generation agent will help sales teams identify potential customers who are actively discussing or seeking solutions in their target market.

We share hands-on tutorials like this 2-3 times a 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 & Connect Your Agents to 100,000+ APIs 🔗📨

Postman, the API platform you likely already use, has launched its AI Agent Builder, bringing powerful AI agent development directly into its ecosystem. The new builder isn't just about connecting to LLMs; it uses Postman's massive API network of 100,000+ APIs to build agents that can discover, connect to, and utilize almost any service.

The platform includes a visual workflow builder, the ability to evaluate and compare different LLMs, and testing tools to ensure your agents perform as expected. This means you can now explore LLMs, integrate with APIs, build complex agent workflows, and test them, all within the familiar Postman environment.

Key Highlights:

1. Unified Workflow - Postman integrates the complete cycle, from exploring APIs and LLMs to building, testing, and deploying AI agents, all within its platform. You can use the Postman client to test LLMs with the same ease and familiarity you use for API endpoints, helping you choose the best model for your specific needs.

2. Visual, No-Code Building with Flows - Postman’s visual editor Flows editor allows you to design complex, multi-step agent workflows using a drag-and-drop interface. Modular flows lets you prototype quickly, and pre-configured templates can be used to start common workflows. For those who prefer working directly with code, Postman provides APIs to generate tools for integrations. This is especially helpful for quickly building the back-end automation your agent will use.

3. Direct Access to a Massive API Ecosystem - Postman’s API network with 100,000+ APIs from 18,000 companies is now directly available to your agents. The system automatically discovers APIs, processes their documentation, and allows the agent to connect to them. Additionally, the new Network Search API enables your agents to programmatically search for specific APIs and datasets.

4. Testing & Evaluation - Postman provides robust testing tools, allowing you to run workflows locally and simulate various scenarios. The platform also enables side-by-side comparison of different LLMs, assessing their speed, token usage, and content quality, helping you make informed decisions.

Complete Opensource Agentic Infrastructure 🤖🏗️

Pica introduces a complete infrastructure stack for building autonomous AI agents, addressing the key challenges we face when implementing agent-based systems. Moving beyond basic API integration, the platform provides tooling for the entire agent development lifecycle - from integration and authentication to execution and workflow management.

At its core is Pica’s OneTool, an SDK that connects agents to 100+ platforms, complemented by AuthKit for secure tool connections, Agent for modular agent development, and AgentFlow for task orchestration. The platform's holistic approach eliminates the need to piece together disparate tools and services.

Key Highlights:

1. End-to-End Development Stack - Get everything needed to build production-ready agents in one place: API integrations via OneTool, secure authentication through AuthKit, modular agent creation with Agent, and multi-agent orchestration using AgentFlow. No need to stitch together multiple services or build custom infrastructure.

2. Integration Layer - OneTool SDK handles schema definitions, authentication flows, and execution processes for 100+ platforms including Salesforce, QuickBooks, and Google Workspace. Write a single line of code to give your agent access to any supported tool, with automatic schema management and authentication handling.

3. Developer-First Security - AuthKit provides a drop-in web component for secure tool connections, letting your users connect their accounts while maintaining granular access controls. Built-in logging and traceability features make debugging and compliance straightforward.

4. Production-Ready Performance - Built on a Rust core for enterprise-grade reliability, the platform handles concurrent agent operations efficiently while guaranteeing memory safety. Scale from prototypes to production without changing your infrastructure.

Quick Bites

Alibaba has released Qwen2.5-Max, a massive MoE (Mixture-of-Experts) language model trained on over 20 trillion tokens and fine-tuned with SFT and RLHF. The model, which outperforms Llama 3.1 405B, GPT-4o, and Claude 3.5 Sonnet in several benchmarks including Arena-Hard and LiveCodeBench, is now available through Alibaba's OpenAI-compatible API interface, allowing you to easily integrate it using familiar OpenAI-style code patterns.

Just after a week of releasing new Sonar APIs, Perplexity AI has integrated DeepSeek R1 reasoning model and introduced Sonar Reasoning, a new API that lets you build products with chain-of-thought reasoning, plus real-time internet search and citations.

Someone just ran full undistilled DeepSeek R1 model locally for around $6,000 using a dual AMD EPYC setup with 768GB DDR5 RAM, achieving 6-8 tokens per second at Q8 quantization without requiring a GPU. The complete build leverages consumer-grade components and a standard tower case, centered around a dual-socket server motherboard to maximize memory bandwidth across 24 RAM channels.

Yohei Nakajima, Adam Silverman, and Alex Reibman have launched Agent Fund, a new VC fund making $50K-$200K investments in AI agent startups across infrastructure, tools, and applications. Operating as an AngelList rolling fund open to accredited investors (with $40K minimum over 4 quarters), the fund aims to build a broad portfolio of companies developing autonomous systems, agent frameworks, and related developer tools.

Tools of the Trade

1. Portals: A drag-and-drop builder for creating custom AI agent workflows that can automate tasks using a library of 30+ connectable nodes (including agents, tools, and triggers). It connects to your documentation, codebases, and other apps to provide context for support and other tasks.

2. OmiAI: LLM SDK for Typescript. It is a highly opinionated AI SDK where you don't pick the model, it auto-picks the best model from a suite of curated models depending on your prompt/messages. It includes built-in reasoning, curated tools, internet access, and full multi-modal support.

3. Emcee: Transforms any web application with an OpenAPI spec into a Model Context Protocol (MCP) endpoint, allowing AI assistants like Claude Desktop to interact with external services and data.

4. Meha: Desktop app that controls your Chrome browser to perform tasks by using LLMs and Playwright. It plans and performs actions based on user prompts, providing a view of the planning steps and access to its long-term memory and file system, and can run multiple agents concurrently.

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

Hot Takes

1. I think the Deepseek moment is not really the Sputnik moment, but more like the Google moment.
   If anyone was around in ~2004, you'll know what I mean, but more on that later… ~
   Yishan
   
   

2. So you are telling me that a Chinese hedge fund released an LLM with UNVERIFIED claims about its training setup and efficiency, which ended up wiping out trillions of dollars in the US stock market, and we are supposed to believe that there is nothing shady going on??? ~
   Bojan Tunguz

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 😉