AI Workflows & Agentic Architecture Directory

pi-multiagent adds isolated multiagent delegation to Pi CLI. Spawn scouts, reviewers, and workers in separate contexts with explicit authority. Step-by-step setup and usage guide.

Build resumable DAG workflows in Pi CLI with pi-taskflow. 18 agents, 8 phase types, cross-session resume, zero dependencies. Complete guide with JSON DSL examples.

Multi-agent swarms are significantly more effective than single AI agents because they solve the 'Context Dilution' problem. By splitting a complex goal into specialized tasks handled by separate agents (e.g., Researcher...

The A2A (Agent-to-Agent) protocol is a horizontal communication standard that allows AI agents from different frameworks to discover, delegate, and collaborate on tasks. Unlike the Model Context Protocol (MCP), which con...

Hermes Agent is winning in 2026 because of its unique 'Skill Crystallization' architecture. Unlike stateless models that require massive prompts for every task, Hermes records its successful tool-use patterns and writes...

Building an A2A-compatible agent involves four steps: installing the A2A SDK, defining your Agent Card (JSON), implementing the task/send handler, and publishing to a well-known endpoint. This allows your agent to be dis...

Microsoft Work IQ API gives AI agents persistent memory across sessions. Boosts cross-session accuracy from 40% to 85%+. Complete setup guide for enterprise developers.

Multi-model tournament code review catches 85-92% of issues before merge. Claude Code dynamic workflows spawn 3-5 competing models. Complete setup guide with cost analysis.

Microsoft MAI-Thinking-1 matches Claude Opus 4.6 on SWE-Bench Pro at a fraction of the cost. Enterprise code review agent with 256K context. Setup guide with pricing.

Google ADK 2.0 ParallelAgent cuts research brief creation from 4 hours to 45 minutes. Open-source multi-agent framework with 3x source coverage. Complete deployment guide.

GitHub Copilot App provides a unified control center for multi-agent development. Worktree isolation, Agent Merge, and multi-tier review. Complete setup and configuration guide.

NVIDIA Nemotron 3.5 Content Safety is a 4B open-weight guardrail model for real-time AI agent moderation. Sub-5ms inference catches nuanced violations keyword filters miss.