Best of ai-agents — September 2025

  1. 1
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·34w

    The Open-source RAG Stack

    A comprehensive guide to building production-ready RAG systems using open-source tools. Covers the complete technology stack from frontend frameworks to data ingestion, including LLM orchestration tools like LangChain and CrewAI, vector databases like Milvus and Chroma, embedding models, and retrieval systems. Also showcases 9 practical MCP (Model Context Protocol) projects for AI engineers, ranging from local MCP clients to voice agents and financial analysts.

  2. 2
    Video
    Avatar of networkchuckNetworkChuck·33w

    N8n AI Agents are INSANE!!

    Demonstrates how to create AI agents in n8n that combine language models with command-line tools and memory. Shows a practical example of building an agent that can check website connectivity by executing ping commands through ChatGPT integration, highlighting the potential for automating network monitoring and system administration tasks.

  3. 3
    Article
    Avatar of bytebytegoByteByteGo·34w

    How Anthropic Built a Multi-Agent Research System

    Anthropic built a multi-agent research system using an orchestrator-worker pattern with a Lead Researcher agent coordinating specialized subagents and a Citation Agent for accuracy. The system outperformed single-agent setups by 90% through parallel processing and dynamic adaptation, though it consumes 15x more tokens. Key engineering principles include proper prompt design, delegation strategies, effort scaling, and parallelization. Production challenges involve managing stateful agents, debugging non-deterministic behavior, and handling deployments without breaking running tasks.

  4. 4
    Article
    Avatar of javarevisitedJavarevisited·36w

    LangGraph and n8n in 2025: The AI Stack You Can’t Ignore?

    LangGraph and n8n serve complementary roles in AI system architecture. n8n excels as a workflow automation tool for connecting APIs, databases, and services, while LangGraph specializes in building intelligent AI agents with multi-step reasoning, state management, and complex tool-calling capabilities. The key insight is using n8n for data movement and integrations, and LangGraph for AI reasoning and agent orchestration, rather than treating them as competing solutions.

  5. 5
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·34w

    Building a Full-stack Agentic App

    AG-UI is an open-source protocol that enables communication between AI agents and frontend interfaces through event-based messaging. The tutorial demonstrates building a stock portfolio agent using CrewAI Flows for the backend and CopilotKit for the frontend, featuring real-time streaming, human-in-the-loop approval, and reactive UI updates. The implementation includes a 5-step workflow covering portfolio initialization, investment parameter extraction, stock data simulation, allocation calculation, and insights generation.

  6. 6
    Video
    Avatar of primeagenThePrimeTime·35w

    LLMs are caught cheating

    AI models like Claude and Qwen Coder were caught using git history to solve coding challenges in the SweetBench benchmark, essentially finding future commits that contained the fixes they needed. While technically cheating, this behavior mirrors real-world software engineering practices where developers search through repository history to understand and fix bugs, especially when backporting fixes to older versions.

  7. 7
    Article
    Avatar of langchainLangChain·36w

    Building LangGraph: Designing an Agent Runtime from first principles

    LangGraph was designed as a low-level agent framework prioritizing production readiness over ease of getting started. Built to address LangChain's feedback about customization and scaling challenges, it focuses on six core features: parallelization, streaming, task queues, checkpointing, human-in-the-loop capabilities, and tracing. The framework uses a structured execution model based on the Pregel algorithm with channels and nodes, enabling deterministic concurrency and fault tolerance. Performance scales gracefully with agent complexity while maintaining low latency, making it suitable for production deployments at companies like LinkedIn, Uber, and Klarna.

  8. 8
    Article
    Avatar of heidloffNiklas Heidloff·34w

    Leveraging Docling in Langflow

    Docling and Langflow are open-source frameworks for document extraction and building AI agents. The integration allows converting PDFs, DOCX, and slides into structured data through Langflow's visual interface. Documents are processed through Docling components, converted to various formats, chunked for semantic search, and used in RAG workflows with vector databases like Astra for question-answering applications.

  9. 9
    Article
    Avatar of tdsTowards Data Science·35w

    How to Build an AI Budget-Planning Optimizer for Your 2026 CAPEX Review: LangGraph, FastAPI, and n8n

    A comprehensive guide to building an AI-powered budget planning system that automates CAPEX portfolio optimization. The solution combines LangGraph for agent orchestration, FastAPI for linear programming optimization, and n8n for workflow automation. Users can send budget requests via email and receive optimized project portfolios that maximize ROI while respecting constraints like sustainability requirements. The system includes email parsing, mathematical optimization using PuLP library, and automated report generation for executive decision-making.

  10. 10
    Article
    Avatar of bytesdevBytes by ui.dev·32w

    Chrome DevTools made an MCP server

    Chrome DevTools released a public preview of their MCP server that enables AI coding agents to control and inspect live Chrome browsers. This tool addresses the limitation of agents coding blindly by providing them with browser inspection capabilities through Puppeteer-powered MCP integration. Agents can now debug applications by inspecting DOM, reading console logs, checking network requests, automating user flows, and running performance audits. The newsletter also covers various web development topics including CSS specificity, TanStack Start migration, Redux usage in 2025, and JavaScript's arguments object handling.

  11. 11
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·32w

    ​6 Popular Agentic Design Patterns Used in AI Products!​

    Explores six key agentic design patterns that power modern AI systems: ReAct (reasoning and action), CodeAct (direct code execution), Modern tool use (external API integration), Self-reflection (error correction), Multi-agent workflows (specialized agent collaboration), and Agentic RAG (intelligent data retrieval). Each pattern addresses specific challenges in AI agent development, with real-world examples from companies like CrewAI, Cursor, and Perplexity demonstrating their practical applications.

  12. 12
    Article
    Avatar of huggingfaceHugging Face·33w

    Gaia2 and ARE: Empowering the community to study agents

    Hugging Face introduces Gaia2, an advanced AI agent benchmark that goes beyond read-only tasks to evaluate interactive behaviors in real-world conditions. Unlike its predecessor GAIA, Gaia2 tests agents on complex scenarios including ambiguity handling, time-sensitive actions, and noise tolerance using a smartphone mock-up environment. The release includes the open-source Agent Research Environments (ARE) framework for running, debugging, and evaluating agents with structured trace recording. Current results show GPT-5 as the top performer, while temporal reasoning remains challenging for all models. The platform enables researchers to create custom scenarios and connect their own tools via MCP integration.

  13. 13
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·33w

    How to Build Agents That Don’t Fail in Production

    Learn to build reliable AI agents for production using Parlant, an open-source framework that prevents costly failures through structured journeys and guidelines. The guide covers creating customer-facing agents with embedded business logic, using real examples like loan approval systems, and includes practical code implementations to ensure agents behave exactly as instructed.

  14. 14
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·35w

    Building a Context Engineering Workflow

    Context engineering involves creating dynamic systems that provide LLMs with the right information, tools, and format to complete tasks effectively. This tutorial demonstrates building a multi-agent research assistant that gathers context from four sources: documents, memory, web search, and ArXiv. The workflow uses Tensorlake for document processing, Milvus for vector storage, Zep for memory management, and Firecrawl for web scraping, orchestrated through CrewAI agents that filter and synthesize responses.

  15. 15
    Article
    Avatar of tdsTowards Data Science·32w

    How to Build Effective Agentic Systems with LangGraph

    LangGraph is an agentic AI framework that simplifies building AI workflows by abstracting state management and tool usage. The framework uses graph-based architecture with nodes (LLM calls or tools) and edges (conditional logic) to create structured workflows. A practical example demonstrates building a document management system with intent classification routing to handle create, delete, and search operations. While LangGraph reduces boilerplate code and provides open-source flexibility, it still requires significant setup and may introduce framework-specific debugging challenges.

  16. 16
    Article
    Avatar of langchainLangChain·35w

    How to turn Claude Code into a domain specific coding agent

    LangChain researchers tested different configurations of Claude Code to improve its performance on domain-specific libraries like LangGraph. They found that high-quality, condensed documentation (Claude.md) combined with tools for accessing detailed information produced the best results. Simply providing raw documentation access didn't improve performance as much as expected and filled up the context window faster. The study used a comprehensive evaluation framework testing functionality, task requirements, and code quality across three different coding tasks.

  17. 17
    Article
    Avatar of ergq3auoeReinier·32w

    Claude Code SDK: The BEST Way to Build AI Agents

    A video tutorial demonstrating how to build AI agents using the Claude Code SDK in under 100 lines of code. Covers creating custom agents with tools, system prompts, MCP servers, streaming input modes, and JavaScript generators for interactive AI behavior control.

  18. 18
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·36w

    [Hands-on] MCP-powered Agentic RAG

    A step-by-step implementation guide for building an Agentic RAG system using the Model Context Protocol (MCP). The system combines vector database search with web search fallback, using Firecrawl for web scraping, Qdrant as the vector database, and Cursor as the MCP client. The tutorial covers setting up an MCP server, creating tools for both vector DB queries and web search, and integrating everything with Cursor for intelligent query routing.

  19. 19
    Article
    Avatar of do_communityDigitalOcean Community·34w

    Getting Started with Agentic AI in LangGraph

    LangGraph is an open-source framework built on LangChain that enables developers to create structured, reliable agentic AI systems using graph-based workflows. Unlike traditional chatbots that provide static responses, agentic AI can reason through problems, make autonomous decisions, and execute multi-step actions. The framework represents AI workflows as nodes (tasks like reasoning or tool usage) connected by edges (decision flows), providing better control and predictability than linear approaches. The tutorial covers setting up LangGraph, building basic agent workflows with state management, and demonstrates practical examples including a calculator agent and a search-enabled reasoning system.

  20. 20
    Article
    Avatar of tdsTowards Data Science·35w

    Building Research Agents for Tech Insights

    A comprehensive guide to building specialized AI research agents that can aggregate and analyze tech content from multiple sources. The approach uses structured workflows, data caching, and prompt chaining to create personalized tech reports. Key components include preprocessing data pipelines, strategic use of small vs large language models for cost optimization, and structured JSON outputs for reliability. The system fetches trending keywords, processes facts from tech forums, and generates themed reports based on user profiles.

  21. 21
    Article
    Avatar of meilisearchMeilisearch·35w

    What is agentic RAG? How it works, benefits, challenges & more

    Agentic RAG enhances traditional retrieval-augmented generation by adding agent-like decision-making capabilities. Instead of static retrieval, it uses planning, validation, and iterative evaluation to determine when and how to retrieve context. The system works through query understanding, retrieval planning, context evaluation, and feedback loops. Key benefits include improved accuracy, better reasoning, and explainability, while drawbacks involve higher computational costs and implementation complexity. Common tools include LangChain, LlamaIndex, and Meilisearch for building these intelligent pipelines.

  22. 22
    Article
    Avatar of googledevsGoogle Developers·35w

    A2A Extensions: Empowering Custom Agent Functionality

    A2A Extensions enable developers to add custom functionalities to agent-to-agent communication beyond the core A2A protocol. Extensions are declared in Agent Cards and identified by unique URIs, creating an open ecosystem. Real-world implementations include traceability extensions for debugging agent interactions, Twilio's latency-aware extensions for voice agents, and Identity Machines' zero-trust handshakes for secure task delegation.

  23. 23
    Article
    Avatar of langchainLangChain·35w

    Agent Middleware

    LangChain 1.0 introduces a new Middleware abstraction to solve the flexibility limitations of traditional agent frameworks. The middleware system allows developers to customize agent behavior through before_model, after_model, and modify_model_request hooks, providing better control over context engineering while maintaining composability. This addresses the common issue where developers abandon agent frameworks for custom code when building production-ready applications.

See all ai-agents archives