AI Workflows & Agentic Architecture Directory

Hy3 beats GLM-5.2 on search and tool orchestration (84.2 BrowseComp, 79.1 MCP-Atlas) with Apache 2.0 license and single-node deployment. GLM-5.2 leads on coding (84.2 SWE-bench Verified) but requires 8x H200 nodes. Hy3 w...

Microsoft Agent Framework (11.9K stars, MIT) is a unified multi-language framework for production multi-agent systems with Python and .NET support, Foundry Hosted Agents, built-in OpenTelemetry, and declarative YAML. Lan...

Webwright is Microsoft Research's browser agent using code-as-action (generates Playwright Python scripts). Browser-Use uses DOM-snapshot with indexed click actions. Stagehand is Browserbase's hybrid NL+code agent. Webwr...

Claude Fable 5 returned to global access on July 1, 2026, after an 18-day suspension under a US export-control directive. The return brings a tighter safety classifier (0.93% prompt-injection success vs 31.5% for Opus 4....

Nanobot is an open-source, ultra-lightweight personal AI agent (45K+ stars, MIT) that deploys via a single pip install. It supports Telegram, Discord, Slack, Feishu, Teams, email, and WebUI channels, with MCP tool integr...

CopilotKit v1.0 is an open-source SDK for building Generative UI in React apps. It provides six UI primitives (Components as Tools, A2UI, MCP Apps, State Rendering, Reasoning, Tool Call Rendering) through the AG-UI Proto...

Microsoft Fara-7B is a 7 billion parameter Computer Use Agent that runs on-device without cloud APIs. It achieves 73.5% on WebVoyager, uses screenshots as input, predicts coordinate-based actions, and costs $1 per trajec...

Microsoft SkillOpt (v0.2.0) is a text-space optimizer that treats agent skill documents as trainable parameters. It uses an optimizer model to propose edits to a skill document, accepts only edits that pass a held-out va...

Otari is Mozilla.ai's open-source LLM control plane with virtual keys, budgets, and built-in tools (code execution, web search). Portkey is an enterprise AI gateway with observability, guardrails, and managed hosting. Li...

By David Mitchell, AI Platform Architect at SaaSNext. I have evaluated 40-plus LLMs for production deployment across 12 enterprise clients and published the Cost-Performance Frontier series on AI model economics. EDITOR...

By Marcus Vance, Lead Performance AI Engineer at SaaSNext. I have built production agent pipelines processing over 2 million requests daily and contributed to the Vercel AI SDK community. I tested WorkflowAgent against r...

OpenScience vs Claude Science compares two 2026 AI workbenches for scientific research. OpenScience is an Apache 2.0, model-agnostic workbench from Synthetic Sciences with 250+ editable skills and 30+ scientific database...