Articles, releases and code from Hacker News, Reddit, GitHub and the people building RPA, workflow automation and AI agents — plus what the community pushed to the top today.
Not enough to go on yet. Open a few links and this list will start leaning towards what you read. 0 of 5 so far.
AILangChain provides an open-source framework and standardized interfaces, letting developers connect language models with external data and tools to build modular AI agent workflows.
AIA curated GitHub repository catalogs Model Context Protocol servers, giving developers prebuilt integrations to connect AI agents with external tools, APIs, and data sources.
AIHeadroom released an open-source tool to compress tool outputs, logs, and RAG context, cutting LLM token usage and costs for AI agents.
AIPrivateGPT released an open-source API layer mirroring Claude's endpoints, allowing developers to build local AI agents and workflows using RAG, MCP connectors, and tools without cloud APIs.
AIAstrBot released an open-source AI agent framework that connects multiple instant messaging platforms with large language models and plugins to automate conversational workflows.
AILangGraph offers an orchestration framework that enables developers to build durable, stateful AI agents featuring human-in-the-loop controls, persistent memory, and fault-tolerant execution.
AIMicrosoft released Playwright MCP, allowing AI agents to control web browsers via structured accessibility snapshots instead of relying on vision models or screenshots.
AIOpenAI released the Python Agents SDK, giving developers a provider-agnostic framework to build multi-agent workflows with built-in tool use, handoffs, sandboxes, and voice pipelines.
AIThe open-source dbx client added built-in AI and an MCP server, enabling AI agents and automated workflows to interact with over 100 database engines.
AIQwen released Qwen3.8-27B, an open-weights vision-language model with native image, video, and computer-use capabilities designed to handle long-horizon agent workflows and coding tasks.
AICrewAI 1.15.40 adds JSONL support for RAG data discovery, enables historical state lookups by row ID, and fixes async execution of after-LLM-call hooks.
AIInngest 1.52.0 adds `step.subscribe` support for workflows to receive events mid-run, alongside a new CLI deploy command and sync history API endpoints.
AIWindmill 1.867.0 enables log streaming for nested subflows and loops, improves AI tool call failure tracking, and fixes Python relative imports across workspace scripts.
AIWindmill 1.875.0 adds fairness reserve settings for worker groups, detailed fairness cap alert timings, and improved run timeline formatting for long-running execution steps.
AITrigger.dev v4.8.5 updates task runtimes and server Docker images to patched Node.js versions, resolving several high-severity security vulnerabilities across HTTP/2 and TLS.
AIWindmill 1.869.0 lets developers run AI decision steps on Gemini's Decisions API and migrate live AI sessions with drafts into workspace forks.
AIn8n 3.3.0 fixes multiple AI agent issues, improving tool-call history retention, custom model ID support, and memory persistence across workflow execution turns.
AIWindmill 1.876.0 lets AI harnesses solve tasks with code and save it for reuse, while passing flow inputs directly to agent tools.
AINode-RED 5.0.10 delivers security fixes, input validation updates, and stability patches to prevent runtime crashes across communications, context storage, and project workflows.
AITrigger.dev updated its documentation with guides for webhook providers, an introduction to AI agents, and instructions for migrating to v3.
AIKestra updated its design system to render workflow topology graphs outside the main Kestra UI, allowing developers to embed flow visualizations into external applications.
AIPrefect released FastMCP, a Python and TypeScript framework that simplifies building Model Context Protocol servers, clients, and interactive tools for AI agents.
AIHaystack provides an open-source Python framework featuring agent lifecycle hooks, concurrent tool calls, and modular pipelines for building customizable, production-ready AI workflows and RAG applications.
AIDapr provides an open-source runtime with durable execution and workflow orchestration APIs, enabling developers to build resilient distributed automations and AI agent systems across environments.
AIThe official Model Context Protocol Python SDK enables developers to build interoperable servers and clients that connect external tools and data sources to LLM-powered agent workflows.
AIA new MCP server connects AI assistants directly to n8n, enabling automated agents to build, test, edit, and debug automation workflows autonomously.
AIThe agency-agents-zh repository provides 277 role-based AI agent prompts and a DAG orchestrator to automate multi-agent workflows across Chinese platforms and developer tools.
AINanobrowser is an open-source browser extension agent that uses local or cloud LLMs to automate tasks directly within your active, logged-in Chrome sessions.
AIBLA released an open-source MCP server and CLI that protects application code, models, and binaries against automated reverse engineering by AI agents.
AIA new Blender add-on introduces an MCP server and multi-agent workflows, enabling developers to automate 3D modeling, scene inspection, and material editing using external LLMs.
AIn8n open-sourced nod, a local personal agent that automates web tasks like chasing invoices and cancelling trials using human-in-the-loop approvals and user-selected AI models.
AIChatWiki released an open-source workflow automation platform for WeChat, enabling developers to build drag-and-drop RAG agents that handle automated messaging, content publishing, and human handoffs.
AIThe open-source Playwright MCP server provides an anti-detect Firefox environment, enabling AI agents to automate web tasks and bypass bot-detection systems without being blocked.
AIAnomaly released browse, an MCP server and Chromium extension that lets AI agents automate actions directly within an existing browser session using saved logins and active tabs.
AIComfy-Org released repackaged model files for MiniMax-H3-Image, enabling developers to run the text-to-image generator directly within local ComfyUI automation workflows.
AIHigurashi AI released Haru-4B, an open-source audio-to-audio foundation model that enables low-latency, full-duplex spoken dialogue for conversational voice agents.
AIThinking Machines released Inkling 2 Small, an open-
AIConcept Country released a modified Qwen-27B model using direct weight programming, improving reasoning benchmark scores while reducing thinking token counts to lower agent inference costs.
AICapri is an open-source, local-first Windows desktop agent that automates files, applications, browsers, and terminals across swappable backends like Claude Code and OpenAI Codex.
AIMeituan open-sourced LongCat-Video-Avatar-1.6, improving lip-sync, motion coherence, and inference speed for automated video generation workflows that create AI avatars exceeding ten minutes.
AINex-AGI released Nex-N2.6, a 321B multimodal model tailored for computer use, featuring improved tool-list adherence and reduced repetitive reasoning for long-horizon agent workflows.
AIMaxwell released Maxwell-V1-Flash, a 35B open-weights MoE model that activates 3B parameters and supports 256K contexts, offering faster, cheaper long-context reasoning for AI agents.
AIMiniMax released an open prompt-expansion model that turns brief user requests into detailed generation prompts with recommended aspect ratios for automated text-to-image workflows.
AIAWS open-sourced UAEF, a framework-agnostic system that lets developers evaluate AI agents across tool calling, reasoning, and safety metrics to automatically detect regressions over time.
AIBIS released MaRCo, an open-source agentic framework that guides AI coding agents through macroeconometric research workflows with automated specification, replication, and human-in-the-loop checkpoints.
AIOpen-source Trirouter routes prompts across Claude Code, Codex, and Antigravity, letting developers automatically control model tiers, parallel agent budgets, and shared skills per task.
AIOpenCode Team provides a configurable, dependency-free multi-agent framework that automates software product lifecycles across fourteen specialized roles with human-in-the-loop approval gates.
AIHirepanel released an open-source multi-agent resume screening system running on local Ollama models with quote verification, prompt-injection defense, and n8n-driven human approval workflows.
AIApache released the first stable Airflow TypeScript SDK, enabling developers to build, trigger, and manage Airflow automation workflows directly within Node.js and TypeScript environments.
AIWindmill 1.872.0 adds background tools for AI agent steps, updates conversation turn handling per memory ID, and allows mounting volumes as read-only without exclusive leases.
AIWindmill 1.873.0 adds a Vertex AI federated resource type for workload identity federation and displays session statuses or unread counts on folded AI session panels.
AIWindmill 1.870.0 adds charts and tables to AI chat replies, checks Git sync PR diffs during polling, and reports failed MCP runs as tool errors.
AIModel Context Protocol servers 2.4.0 patches six security vulnerabilities across filesystem, Git, and fetch integrations, preventing symlink directory escapes and unauthorized remote command execution in automated agent workflows.
AIKestra updated its topology package to let developers render workflow graphs outside the main Kestra user interface.
AIWindmill 1.868.0 compiles Tailwind CSS v4 in app bundles, restores structured output tools for Claude, and now treats skipped jobs as successful in wait paths.
AITrigger.dev updated its documentation to detail personal access tokens, CLI login expiry, and trusted publishing configurations for automated workflows.
AIPlaywright 1.65.1 fixes a regression affecting decorated classes with static fields and corrects test location offsets when running browser automation under Bun.
AIWindmill 1.873.1 fixes dedicated flow runner routing, preserves static input references in app deployments, and improves worker error context for debugging automations.
AIPrefect 3.9.3 fixes early closures in async generator flows, resolves parallel task failures with keyword arguments, and stops parent transactions from committing after child failures.
AITrigger.dev now logs the underlying causes of retryable alert delivery failures, helping developers diagnose and debug workflow notification issues more easily.
AILangGraph 1.2.21 fixes subgraph namespace matching and reading issues in multi-agent workflows while updating dependencies for security.
AICrewAI 1.15.41 adds support for approved private URL hosts and preserves earlier step outputs when resuming workflows after human feedback.
AIWindmill 1.871.1 fixes a dangling symlink issue that caused release build failures, ensuring reliable deployment builds for self-hosted automation infrastructure.