Agent Infrastructure

• AgentField - 2.0 | 2026-10 | Agent identity and RPC using W3C DIDs |
• Mem0 - 2.0 | 2026-10 | Memory layer that persists across agent sessions |
• openma - 2.0 | 2026-10 | Self-hosted Managed Agents API implementation |

Autonomous Agents (2023 wave)

• BabyAGI - 01 | Minimal task-driven autonomous agent loop |
• XAgent - 2.0 | 2026-07 | Autonomous agent with planning and tool learning |
• OpenManus - 09 | General-purpose agent, no invite code needed |
• AutoGPT - 10 | The original autonomous GPT-4 agent loop |

CLI Agent Harnesses

• Atomic Agent - 10 | Local-first CLI agent for open-weight models |
• Agent Teams - 3.0 | 2026-10 | Desktop app running coding-agent teams across CLIs |
• Bernstein - 2.0 | 2026-10 | Deterministic orchestrator for 40+ CLI agents |
• SwarmClaw - 06 | Self-hosted runtime for multi-agent CLI work |
• h5i - 2.0 | 2026-10 | Runs agents in sandboxes, merges the verified result |
• Dorothy - 07 | Desktop app to run several CLI agents at once |
• ClawFleet - 04 | Deploys isolated agent instances via Docker |
• OpenPaw - 05 | Turns Claude Code into an assistant with 38 skills |
• ORCH - 08 | One CLI to manage a team of agents on tasks |
• OpenHermit - 10 | Deploys agent fleets as long-running services |
• 5dive - 10 | Run a company of named agents on your own server |
• Agon - 10 | Claude Code plugin for autonomous research loops |
• TeamHero - 04 | Manage agents like a team, with structured roles |
• OpenSepia - 03 | Nine Claude agents running as an agile team |

Core Frameworks

• Semantic Kernel - 10 | Plugin-based AI integration for .NET and Python |
• AGiXT - 07 | Multi-provider agent platform with command chaining |
• ix - 10 | Autonomous agents with a visual workflow builder |
• Smolagents - 2.0 | 2026-10 | Minimal agents that write code to act |
• Pydantic AI - 10 | Type-safe agents on Pydantic with structured output |
• Upsonic - 06 | Agents with MCP and isolated execution |
• Google ADK - 2.0 | 2026-10 | Code-first agents that deploy to Vertex AI |
• Neurolink - 10 | One interface across 12+ LLM providers |
• Mastra - 10 | TypeScript agents with RAG and observability |
• Axar - 2.0 | 2026-02 | Minimal TypeScript agents with Zod validation |
• Tambo - 10 | React components rendered by AI at runtime |
• OpenClaw - 10 | Personal AI assistant that runs on any platform |
• LangChain - 10 | Compose LLM apps from modular pieces |
• Openwork - 08 | Open-source AI coworker platform |
• Hive - 2.0 | 2026-10 | Multi-agent harness aimed at production |
• Atomic Agents - 10 | Compose agents from small interchangeable parts |
• OpenAgent - 2.0 | 2026-10 | Personal assistant built on LLM, RAG and agent loops |
• AG2 - 2.0 | 2026-10 | Community fork of AutoGen, now an AgentOS |
• trpc-agent-go - 2.0 | 2026-10 | Go framework for agents with graph workflows |
• ConnectOnion - 2.0 | 2026-10 | Python framework focused on agent collaboration |
• LightAgent - 2.0 | 2026-09 | Lightweight Python agents with tools and memory |
• Agentlas OS - 2.0 | 2026-10 | Specialist agent hub with temporary orchestrators |
• Ouroboros - 10 | Agent runtime with reviewed self-modification |
• Oh My Hermes - 10 | Harness with optimized tools and memory |
• Promptise Foundry - 2.0 | 2026-10 | Agentic framework with controllable reasoning |
• Aeon - 10 | Runs unattended on GitHub Actions, self-healing |
• Octochains - 08 | Parallel isolated reasoning with an aggregator |
• OpenProgram
• Octomind - 2.0 | 2026-10 | Model-agnostic runtime with specialist agents |
• ProtoLink - 10 | Python agents with native A2A communication |
• NarraNexus - 2.0 | 2026-10 | Builds nexuses where agent intelligence emerges |
• ShaprAI
• TrashClaw - dependency local agent for old hardware |
• KodeAgent - 2.0 | 2026-10 | Minimal agent engine, deliberately small |

Domain-Specific Agents

• GenoMAS - 04 | Multi-agent pipeline for genomics data analysis |
• CleverBee - 3.0 | 2026-01 | Deep research agent that browses with Playwright |
• RAI - 2.0 | 2026-10 | Agent framework for robotics, built on ROS 2 |
• DeepAnalyze - 10 | Autonomous data science without fixed workflows |
• Darkmoon - 3.0 | 2026-10 | Autonomous pentesting across web, cloud and AD |
• text2sql-framework - 08 | Text-to-SQL agent that explores schema, not RAG |
• wechat-mac-rpa - 10 | Visual agent automating WeChat on macOS |
• Omni-Rewriter
• DNA Claude Analysis - 03 | Explore your genome in natural language |
• everyrow - 10 | Run LLM agents over pandas DataFrames |
• Inalpha - 3.0 | 2026-10 | Quant agents that pick factors that still work |

Inactive

• SuperAGI - 01 | Autonomous agent platform with a tool framework |
• OpenAgents (XLang) - 2.0 | 2024-11 | Platform for data, web and coding agents |
• AI Legion - 05 | TypeScript swarm of autonomous agents |
• Flowise - 08 (archived) | Drag-and-drop builder for LLM flows |
• Agent Protocol - 04 | Standard interface for agent interoperability |

Low-Code & Visual Builders

• Dify - 10 | Visual orchestration for LLM apps and agents |
• Kiln AI - 10 | Desktop app for evals, RAG and fine-tuning |
• Heym - 10 | Visual builder for agentic workflow automation |

Memory & Context

• Hindsight - 10 | Agent memory with retain, recall and reflect |
• Caura - 2.0 | 2026-10 | Governed shared memory for fleets of agents |
• AnimaWorks - 2.0 | 2026-10 | Organization-as-code with brain-inspired memory |
• OMEGA - 2.0 | 2026-10 | Persistent memory for coding agents over MCP |
• Perseus
• Inite Brain - 3.0 | 2026-10 | Bitemporal knowledge graph as agent memory |

Multi-Agent Orchestration

• AutoGen - BY-4.0 | 2026-10 | Conversational multi-agent systems |
• CAMEL - 2.0 | 2026-10 | Role-playing agents for studying agent society |
• Flock - 10 | Declarative agents via blackboard architecture |
• hcom - 10 | Agents message and spawn each other in terminals |
• PraisonAI - 10 | Multi-agent workflows with self-reflection |
• OpenAgents - 2.0 | 2026-10 | Agent networks over WebSocket, gRPC, MCP and A2A |
• Quorum - 09 | Structured multi-agent debate in the terminal |
• auto-co - 06 | 14 agents run a company in a continuous loop |
• MetaGPT - 01 | Agents role-play a software company |
• CrewAI - 10 | Orchestrate role-playing agent crews |
• Markus - 2.0 | 2026-10 | Agents coordinate and review each other's work |
• CommonGround Kernel - 2.0 | 2026-05 | Postgres-backed shared substrate for agent teams |
• OpenAcme - 07 | Role-specialized agents that self-organize |
• Hivekeep - 10 | Self-hosted team of persistent personal agents |

Research & Experimental

• EvoAgentX - 08 | Agent workflows that evolve and self-optimize |
• AgentFlow - 02 | Trainable multi-agent system using Flow-GRPO |
• AgentSquare - 11 | Automatic search over modular agent designs |
• Cache-to-Cache - 2.0 | 2026-10 | Agents exchange meaning directly via KV-cache |
• agent-opt - 2.0 | 2026-06 (archived) | Optimizes prompts and agent workflows |
• GNAP - 03 | Git-native protocol draft for agent coordination |
• AVP - 2.0 | 2026-04 | Transfers KV-cache between agents, not text |

Retrieval & Data

• LlamaIndex - 10 | Connects LLMs to 160+ data sources |
• Haystack - 2.0 | 2026-10 | Composable pipelines for search and RAG |

Safety, Security & Evaluation

• Agentic Radar - 2.0 | 2025-11 | Scans agent workflows for CVE and OWASP issues |
• Cordum - 10 | Evaluates policy before an agent action dispatches |
• Greywall - 2.0 | 2026-10 | Deny-by-default sandbox for coding agents |
• Kitaru - 2.0 | 2026-10 | Record, replay and improve agents in production |
• RapidFire AI - 2.0 | 2026-10 | Experiment harness for RAG and fine-tuning runs |
• APort Guardrails - 10 | Pre-action authorization policy for agent calls |