A living encyclopedia of agent frameworks tools — independently surveyed, benchmark-ready, and continuously updated.
Authored by Team Ardur · CC BY 4.0
| Tool | Type | License | Tier | Notes |
|---|---|---|---|---|
| Agno | Full-stack | Apache-2.0 | A | 39.7K+ stars; multi-agent + memory; AgentOS runtime; multimodal |
| AutoGen / AG2 | Multi-agent | MIT | A | 43.1K+ stars; AutoGen v0.7 = research; AG2 = community fork; MAF = successor |
| Claude Agent SDK | SDK | Open SDK | A | Fastest-growing; Anthropic-native; MCP; skills; subagents |
| Coze | No-code | Proprietary | A | ByteDance; 100K+ bots; zero-code Bot builder; Doubao model |
| CrewAI | Multi-agent | MIT | A | 49.2K+ stars; 63% Fortune 500 claim; role-based crews; MCP support |
| Deep Agents | Agent Harness | MIT | A | LangChain ecosystem; long-running task harness; planning + context mgmt |
| Dify | Low-code | Apache-like | A | 142K+ stars; workflow + RAG + agent; prompt IDE; self-hostable |
| DSPy | Programming | Open | A | Stanford; programming (not prompting) LM systems; prompt + weight optimization |
| Flowise | Low-code | Apache-2.0 | A | 54.9K+ stars; LangChain DAG on canvas; LangGraph nodes v2.0 |
| Google ADK | Framework | Apache-2.0 | A | 15.6K+ stars; hierarchical agent tree; A2A + multimodal; Vertex AI |
| Google Vertex AI Agent Builder | Enterprise | Proprietary | A | Agent Studio (no-code) + ADK (code); Agent Runtime; BYO-MCP |
| Haystack | Pipeline | Open | A | 18K+ stars; modular retrievers, routers, evaluators; enterprise doc intelligence |
| LangChain | Framework | MIT | A | 134K+ stars; 1000+ connectors; integration layer; increasingly tooling |
| LangGraph | Orchestration | MIT | A | 44% prod adoption; graph-based state machines; checkpointing; HITL |
| Letta | Stateful Agent | Open | A | 22.3K+ stars; memory-rich agents; state/session management |
| LlamaIndex | RAG/Agent | MIT | A | 40.9K+ stars; data-first; Router Agents; retrieval-centric |
| Mastra | Framework | Apache-2.0 | A | 24.6K+ stars; 300K+ weekly npm; TS-native; 3,300+ models, 94 providers |
| Microsoft Agent Framework | Unified SDK | MIT | A | 9.6K+ stars; GA April 2026; merges AutoGen + Semantic Kernel; MCP + A2A native |
| Microsoft Copilot Studio | Enterprise | Proprietary | A | M365-integrated; low-code; credit model; Power Platform |
| n8n | Automation | Fair-code | A | 188K+ stars; 400+ integrations; AI Agent node; fair-code (not OSI) |
| OpenAI Agents SDK | SDK | MIT | A | 22.2K+ stars; lightweight; explicit handoffs; built-in tracing |
| Pydantic AI | Framework | Open | A | 8.4K+ stars; type-safe; DI; FastAPI ergonomics; 5 output modes |
| Salesforce Agentforce | Enterprise | Proprietary | A | CRM-native; $2/conversation; Atlas Reasoning Engine; multi-agent |
| ServiceNow Now Assist | Enterprise | Proprietary | A | ITSM-native; CMDB grounding; orchestrator model |
| smolagents | Minimalist | Open | A | HuggingFace; ~1K LOC core; CodeAgent writes/executes Python; GAIA 44.2% |
| Vercel AI SDK | SDK | Open | A | Low-level primitives; Mastra builds on top; model routing + streaming |
| Adobe AEP Agent Orchestrator | Enterprise | Proprietary | B | Marketing; personalization agents; real-time data |
| AgentScope Java | Multi-agent | Open | B | Alibaba; native A2A protocol; Hook system for HITL; Qwen-first |
| Airia | Enterprise | Proprietary | B | Award 2026; unified security + orchestration + governance |
| ArkClaw | Enterprise | Proprietary | B | ByteDance; Volcano Engine; enterprise process automation |
| AstronClaw | Voice Agent | Proprietary | B | iFlytek; 98%+ ASR; dialect support; streaming |
| AutoAgents | Multi-agent | MIT | B | ~2K stars; Actor model (Ractor); WASM sandbox; derive macros |
| Botpress | Conversational | Open | B | Open-source dialog + LLM; modular; community extensible |
| Chapter Enterprise | Enterprise | Proprietary | B | Model-agnostic; SmartGuard; legacy on-premise + SaaS; air-gapped |
| Domo Agent Catalyst | Enterprise | Proprietary | B | BI-native; governed datasets; multiple LLM choice; HITL |
| Embabel | Framework | Open | B | Rod Johnson (Spring creator); GOAP + Utility AI planning; deterministic; beta |
| Genkit | Framework | Open | B | Firebase/Google; Dev UI with trace explorer; typed flows; OTEL-compatible |
| Harness AI Agent Framework | Enterprise | Proprietary | B | SDLC-wide; 30-50% productivity gain; AIDA chat; MCP + AIDA |
| Hermes Agent Framework | Framework | Open | B | 106K+ stars; Workspace + Mission Control; Obsidian memory; 24/7 scheduled |
| IBM watsonx Orchestrate | Enterprise | Proprietary | B | AI orchestration; enterprise governance; hybrid cloud |
| kagent | Framework | Open | B | Google-backed; Kubernetes-native; 45ms latency |
| Kimi Claw | Long-context | Proprietary | B | Moonshot; 2M-character context; legal/academic/financial |
| Koog | Framework | Open | B | JetBrains; graph workflows; checkpointing; Kotlin Multiplatform; history compres |
| LangChain4j | Framework | Open | B | 30+ vector stores; 20+ providers; Quarkus/Micronaut; AI Services pattern |
| Langflow | Visual | Open | B | 54.9K+ stars; visual LangChain builder; node-based; Git versionable |
| MaxClaw | Multi-modal | Proprietary | B | MiniMax; MoE architecture; text/voice/video/image |
| MoltBook | Open Templates | Apache-2.0 | B | ByteDance; 15K+ stars; OpenClaw-compatible; Skill library |
| OpenClaw | Self-hosted | MIT | B | 210K+ stars; self-hosted; 50+ native integrations; Docker-first |
| OpenFANG | Agent OS | Apache-2.0 | B | ~8K stars; 137K LOC; kernel/runtime/memory/channels; RBAC; MCP + A2A |
| QwenPaw | Agent | Proprietary | B | Alibaba; e-commerce/service automation; agent store |
| RAGflow | RAG Engine | Open | B | 80.7K+ stars; deep document understanding; complex PDF/table parsing |
| Rasa | Conversational | Open | B | 20K+ stars; 25M+ downloads; open-source + Rasa Studio |
| Relevance AI | No-code | Proprietary | B | Drag-and-drop; zero coding; non-technical users |
| Rig | LLM Lib | Apache-2.0 | B | ~5K stars; trait-first; composable; CompletionModel + Tool traits |
| Semantic Kernel | SDK | MIT | B | Pre-MAF; plugin/planner model; .NET enterprise; KPMG/Fujitsu production |
| Spring AI | Framework | Apache-2.0 | B | Spring-native; Advisors API; auto-configuration; Actuator observability |
| UiPath Agentic Automation | Enterprise | Proprietary | B | RPA-native; visual orchestration + conversational UI |
| Voiceflow | No-code | Proprietary | B | Visual agent builder; voice + chat; enterprise teams |
| 文心智能体平台 | Platform | Proprietary | B | Baidu; 100K+ agents; vertical agents (legal, medical); Wenxin 5.0 |
| 智谱 AutoClaw | Framework | Open (GLM) | B | Zhipu; GLM-4; 256K context; task planning; self-correction |
| Appsmith | Agent Builder | Open | C | Custom AI agents for sales/support/people management |
| Bland AI | Voice | Proprietary | C | High-volume outbound; webhook CRM |
| BUMBLE | Robotics | Open | C | Building-wide mobile manipulation; VLM-based; skill library |
| CodeBuddy | Coding | Proprietary | C | Tencent; AI programming assistant (2026 new) |
| Cognosys | Agent Builder | Proprietary | C | Web-based agent builder; task automation |
| EchoLeads | Voice Sales | Proprietary | C | Startup sales voice AI; 72h deploy; Salesforce/HubSpot |
| Instructor | Structured Output | Open | C | 11K+ stars; patching LLM clients for guaranteed Pydantic output |
| JVSClaw | Low-code | Proprietary | C | Alibaba; low-code AI agent platform |
| Kore.ai | Conversational | Proprietary | C | Healthcare/finance/retail; NLP + automation; HIPAA compliance |
| LeRobot | Robotics | Open | C | HuggingFace; OpenVLA; 7B open-source; data collection + policy training |
| MetaGPT | Multi-agent | Open | C | Software company simulation; roles: PM/tech-lead/engineer; ICLR 2024 oral |
| OrchestrAI | AI OS | Proprietary | C | Non-technical ops; modular capacity; auto-learning from feedback |
| QoderWork | Coding Agent | Proprietary | C | Alibaba; AI programming workflow agent |
| RAI | Robotics | Open | C | Multi-agent embodied; ROS2; perception/planner/controller |
| Retell AI | Voice | Proprietary | C | Custom voice AI; streaming APIs; technical teams |
| ROSA | Robotics | Open | C | ROS-based LLM agent; LangChain + ReAct; JPL NeBula-Spot |
| SuperAGI | Autonomous | Open | C | Marketplace of tools/skills; monitoring dashboards; self-improving |
| Synthflow | Voice | Proprietary | C | No-code voice; 30-min setup; non-technical; $29-450/mo |
| TARS | Embodied AI | Proprietary | C | AWE 3.0 foundation model; DexHand; industrial robots |
| Vapi | Voice | Proprietary | C | Developer API; max customization; engineering-led |
| WorkBuddy | Enterprise | Proprietary | C | Tencent; enterprise collaboration AI agent |
| 红手指Operator | Browser Agent | Proprietary | C | Baidu; browser automation; operator-style task execution |
- OrchestrAI (Tier C)
- QwenPaw (Tier B)
- Deep Agents (Tier A)
- OpenFANG (Tier B)
- n8n (Tier A)
- SuperAGI (Tier C)
- 红手指Operator (Tier C)
- CodeBuddy (Tier C)
- QoderWork (Tier C)
- TARS (Tier C)
- Google Vertex AI Agent Builder (Tier A)
- Microsoft Copilot Studio (Tier A)
- Salesforce Agentforce (Tier A)
- ServiceNow Now Assist (Tier A)
- Adobe AEP Agent Orchestrator (Tier B)
- Airia (Tier B)
- ArkClaw (Tier B)
- Chapter Enterprise (Tier B)
- Domo Agent Catalyst (Tier B)
- Harness AI Agent Framework (Tier B)
- IBM watsonx Orchestrate (Tier B)
- UiPath Agentic Automation (Tier B)
- WorkBuddy (Tier C)
- Google ADK (Tier A)
- LangChain (Tier A)
- Mastra (Tier A)
- Pydantic AI (Tier A)
- Embabel (Tier B)
- Genkit (Tier B)
- Hermes Agent Framework (Tier B)
- kagent (Tier B)
- Koog (Tier B)
- LangChain4j (Tier B)
- Spring AI (Tier B)
- 智谱 AutoClaw (Tier B)
- Agno (Tier A)
- Rig (Tier B)
- Kimi Claw (Tier B)
- smolagents (Tier A)
- AutoGen / AG2 (Tier A)
- CrewAI (Tier A)
- AgentScope Java (Tier B)
- AutoAgents (Tier B)
- MetaGPT (Tier C)
- MaxClaw (Tier B)
- Coze (Tier A)
- Relevance AI (Tier B)
- Voiceflow (Tier B)
- MoltBook (Tier B)
- LangGraph (Tier A)
- Haystack (Tier A)
- 文心智能体平台 (Tier B)
- DSPy (Tier A)
- RAGflow (Tier B)
- LlamaIndex (Tier A)
- Claude Agent SDK (Tier A)
- OpenAI Agents SDK (Tier A)
- Vercel AI SDK (Tier A)
- Semantic Kernel (Tier B)
- OpenClaw (Tier B)
- Letta (Tier A)
- Instructor (Tier C)
- Microsoft Agent Framework (Tier A)
- Langflow (Tier B)
- AstronClaw (Tier B)
- EchoLeads (Tier C)
- Browse tools — Start with the Tier A tools above; they are the most production-ready.
- Read the methodology — See
methodology/for how tools are evaluated and benchmarked. - Check editions — Quarterly editions provide curated analysis and trend reports.
- Contribute — See
CONTRIBUTING.mdfor how to add or update tools.
Latest: 2026-06.md
Tools are evaluated across multiple dimensions including accuracy, latency, token economics, scale behavior, ops burden, developer experience, and data sovereignty. See methodology/ for details.
See CONTRIBUTING.md for guidelines on adding tools, updating analyses, and submitting benchmarks.
Content: CC BY 4.0 — Authored by Team Ardur Code/Benchmarks: As noted in individual files