-2
The Top 100 AI Agent Harnesses by Estimated Popularity (reddthat.com)
submitted 1 month ago by to c/PumpkinDrama@reddthat.com
1 comment fedilink hide all child comments
 

Based on current data from GitHub stars, community momentum, and industry adoption, here is a ranked list of the 100 most popular AI agent harnesses. The ranking prioritizes agent harnesses—the runtimes and frameworks that enable LLMs to execute tools in a loop—over pure orchestration libraries where possible, though the two categories overlap significantly.

The Top 100 AI Agent Harnesses by Estimated Popularity

Rank Harness Est. Stars Category Primary Language Notes
1 OpenClaw 389k Personal Agent Runtime TypeScript Multi-channel gateway (Telegram, Discord, Slack, WhatsApp); most mature personal agent harness
2 Superpowers 283k Coding Harness Config Shell Agentic skills framework & dev methodology
3 mattpocock/skills 258k Skills Collection — Top weekly star gainer; agent skills ecosystem
4 Hermes Agent 244k Personal Agent Runtime Python Three-layer persistent memory; "the agent that grows with you"
5 DeepSeek Harness 217k Harness Runtime Python/TS "Everything-is-a-plugin" design; fastest GitHub star record (141k in 4 days)
6 OpenCode 206k Coding Agent TypeScript Terminal-first coding harness with broad LLM provider coverage
7 n8n 203k Workflow Platform TypeScript Visual workflow automation with AI steps
8 AutoGPT 187k Autonomous Agent Python Original autonomous agent framework
9 Anthropic Skills 173k Skills Collection — Official Anthropic public agent skills repo
10 Langflow 154k Low-Code Platform Python Visual AI agent builder
11 Dify 154k Low-Code Platform Python Most-adopted visual AI agent builder
12 LangChain 146k Framework Python/JS Largest ecosystem for LLM app composition
13 Claude Agent SDK 144k Coding Harness Python/TS Deepest MCP integration; subagents; OS access
14 GStack 130k Coding Harness Config TypeScript Skills/config collection
15 Codex CLI 123k Coding Agent Python OpenAI's terminal coding agent; sandboxed execution
16 Browser Use 114k Web Agent Python Web automation agent harness
17 Gemini CLI 107k Coding Agent TypeScript Google's first-party terminal agent for Gemini
18 Pi 103k Coding Agent Rust Terminal-first coding harness; broad LLM coverage
19 TradingAgents 104k Research Harness Python Multi-agent trading analysis
20 claude-mem 93k Memory Layer — Persistent memory for Claude agents
21 addyosmani/agent-skills 91k Coding Harness Config — Skills collection for agent workflows
22 MCP Servers 90k Plugin/MCP TypeScript Official Model Context Protocol server collection
23 OpenHands 86k Coding Agent Python Full open-source SWE agent (CLI + GUI + cloud)
24 DeerFlow 81k Research Harness Python ByteDance's long-horizon SuperAgent harness
25 Ruflo 72k Agent Meta-Harness TypeScript "The original agent meta-harness"
26 CrewAI 58k Multi-Agent Framework Python Role-based multi-agent orchestration; ~2B agent executions
27 LlamaIndex 52k Data Agent Framework Python RAG-heavy agent workflows
28 Agno 42k Multi-Agent Runtime Python High-performance agent framework
29 LangGraph 41k Orchestration Python/TS Graph-based stateful multi-agent workflows
30 DSPy 38k Optimization Python Programming model for optimizing LLM pipelines
31 SWE-agent 36k Coding Harness Python Original SWE-bench agent
32 AutoGen (Microsoft) 35k Multi-Agent Framework Python/.NET Conversational multi-agent coordination
33 Semantic Kernel 29k Enterprise Framework Python/.NET Microsoft's enterprise agent SDK
34 OpenAI Agents SDK 29k Orchestration Python/TS Clean multi-agent delegation; minimal abstraction
35 Mastra 28k TypeScript Framework TypeScript TS-first with workflows, memory, and Studio
36 Haystack 28k RAG/Agent Framework Python Enterprise-grade retrieval and agents
37 Google ADK 22k Orchestration Python/Java/Go Opinionated, batteries-included agent runtime
38 Pydantic AI 20k Structured Agent Framework Python Type-safe agents with validators
39 OpenHarness 15k Safety Harness Python Safety layer for LLM agents
40 Cline 14k Coding Agent TypeScript VS Code extension agent
41 Aider 14k Coding Harness Python Git-aware context; diff-based editing
42 Roo Code 12k Coding Agent TypeScript Cline-lineage VS Code/Cursor extension
43 Composio 30k Tooling Python/JS 250+ tool integrations for AI agents
44 Mem0 12k Memory Layer Python Drop-in memory for agents
45 AG2 (AutoGen 2.0) 5k Multi-Agent Framework Python Successor to AutoGen
46 Smolagents 18k Lightweight Framework Python HuggingFace's minimalist agent framework
47 MetaGPT 15k Multi-Agent Framework Python SOP-driven role pipeline
48 Vercel AI SDK 12k TypeScript SDK TypeScript Streaming-first AI SDK
49 Goose 11k Coding Agent — BYOM (Bring Your Own Model) agent
50 Agent Swarm 10k Multi-Agent Orchestration — Orchestration for Claude Code, Codex, Gemini CLI
51 MCP-Zero 9k Tool Management — Plug hundreds of tools without context bloat
52 ToolGen 8k Tool Management — ICLR 2025 tool generation
53 ToolRAG 8k Tool Management — Retrieval-augmented tool selection
54 Rasa 7k Conversational Agent Python Business-rule-constrained conversational agents
55 SWE-ReX 6k Execution Sandbox Python Sandboxed execution for SWE agents
56 E2B 6k Execution Sandbox — Purpose-built sandbox for AI agents
57 Modal 5k Execution Runtime Python Serverless execution for agent workloads
58 Blaxel 4k Agent Hosting — Sandboxes, agents hosting, MCP servers
59 Daytona 4k Dev Environment — Development environment infrastructure
60 Trigger.dev 4k Workflow Engine TypeScript Background jobs for agent workflows
61 Conductor 3k Orchestration — Event-driven agent orchestration
62 Sim 3k Agent Builder — Self-hostable agent workspace
63 Flowise 3k Low-Code Platform JavaScript Visual AI agent builder
64 AgentGPT 3k Autonomous Agent TypeScript Browser-based autonomous agents
65 BabyAGI 3k Autonomous Agent Python Task-driven autonomous agent
66 SuperAGI 3k Autonomous Agent Python Open-source autonomous AI agent framework
67 AgentScope Runtime 2k Agent Runtime Python Agent runtime from Alibaba
68 SWE-ReX 2k Execution Sandbox Python Sandboxed execution for SWE agents
69 sandboxed.sh 2k Execution Sandbox — Shell-based sandboxing
70 HexAgent 2k Agent Runtime — UnicomAI's agent harness
71 LangGraph BigTool 2k Tool Management Python Tool management at scale
72 Graphiti (Zep) 2k Memory Layer Python Temporal knowledge graph memory
73 agentlog 2k Memory Layer — Agent logging and memory
74 PraisonAI 2k Multi-Agent Python Multi-agent orchestration
75 agent-squad 2k Multi-Agent Python AWS multi-agent orchestration
76 Microsoft Agent Framework 2k Enterprise Framework Python/.NET Successor to AutoGen + Semantic Kernel; 1.0 GA April 2026
77 AWS Strands Agents 1k Model-Driven SDK Python AWS's model-driven agent SDK
78 JCode 1k Coding Harness Rust Fast Rust-based coding harness
79 Omnigent 1k Agent Runtime Python Claude Code-compatible runtime
80 AgentField 1k Control Plane — Makes AI agents callable as microservices
81 CopilotKit 1k Agent UI TypeScript In-app AI copilot framework
82 KaibanJS 1k Multi-Agent TypeScript Kanban-style agent orchestration
83 ControlFlow 1k Workflow Engine Python Prefect's agent workflow engine
84 L2MAC 1k Computer Framework — LLM Automatic Computer Framework
85 Patchwork 1k Enterprise Workflow — Agentic AI for enterprise workflow automation
86 Swarms 1k Multi-Agent Python Swarm intelligence framework
87 TaskWeaver 1k Code-First Agent Python Microsoft's code-first agent framework
88 BeeAI 1k Agent Framework — IBM's agent framework
89 Any-Agent 1k Framework Evaluator — Mozilla's interface to evaluate agent frameworks
90 LangChain BigTool 1k Tool Management Python Tool retrieval for LangGraph
91 Firecracker 1k MicroVM Sandbox Rust Lightweight VM for agent sandboxing
92 Northflank 1k Agent Runtime — Container PaaS for agent workloads
93 Cloudflare Workers 1k Edge Runtime JavaScript V8 isolate-based edge execution
94 Fly.io Machines 1k Agent Runtime — Infrastructure for agent deployment
95 Agentlas OS 1k Agent OS — Agent operating system
96 DongTai-agent-java 700 Java Agent Java Java application probe
97 wanman 681 Agent Runtime TypeScript Self-hosted agent runtime
98 arcjet-js 681 Security TypeScript Agent security layer
99 swarmclaw 654 Swarm Runtime TypeScript Open-source self-hosted agent swarm runtime
100 jcode 500 Coding Harness Rust Fast Rust coding harness

Key Observations

The harness era is real. The top of the list is dominated by purpose-built agent runtimes (OpenClaw, Hermes Agent, DeepSeek Harness, Claude Agent SDK) rather than traditional orchestration frameworks. This reflects the field's convergence on the insight that harness quality, not just model quality, determines whether agents ship.

Star velocity tells a different story than cumulative stars. DeepSeek Harness crossed 141k stars in four days—the fastest GitHub star record for any AI project. Hermes Agent hit 90k in six weeks. Meanwhile, Browser Use leads the Pulse Score (momentum-based) leaderboard with a 90.3 score, driven by +1,109 weekly stars.

The market is segmenting into three distinct layers: (1) Harness runtimes that close the loop between model and tools (OpenClaw, DeepSeek Harness, Claude Code), (2) Orchestration frameworks for multi-agent workflows (LangGraph, CrewAI, AutoGen), and (3) Low-code platforms for business teams (Dify, Coze, n8n).

MCP is becoming the standard interface layer. Nearly every major harness now supports the Model Context Protocol, with 200+ server implementations available. Tool management harnesses (MCP-Zero, ToolGen, ToolRAG) are emerging specifically to solve the context-bloat problem of scaling to hundreds of tools.

Note on methodology: Star counts are drawn from multiple leaderboards as of September 2026 and represent estimates where sources differ. The ranking blends cumulative stars with velocity metrics (Pulse Score, weekly gains) to avoid over-weighting older projects that accumulated stars before the current agentic boom. Some projects (e.g., LangChain at #12) are technically orchestration libraries rather than pure harnesses, but are included because they are commonly used as harness foundations.

you are viewing a single comment's thread
view the rest of the comments
[–] [S] 1 point 1 month ago

The agent harness landscape has a long tail of genuinely well-architected projects that haven't yet accumulated the stars of OpenClaw or DeepSeek Harness. Below are 20 hidden gems — each with a distinct architectural idea worth studying, and each with a star count low enough that it still feels like a discovery.

# Harness Category Language Est. Stars Why It's a Hidden Gem Key Design Strength
1 SimonAgent Agent Runtime Python ~200 "Zero frameworks, zero magic" — built from scratch in pure Python with a small, auditable core. 73 unit tests cover the loop, context, memory, and security branches. Dual-layer context compression (local sliding window + LLM compaction), on-demand skill loading, and a security boundary that blocks arbitrary command execution and restricts web access to public addresses.
2 DragonScale Embedded Agent Runtime Rust/Go ~500 A compact runtime for Linux and embedded environments that fixes the architectural gaps in single-binary agents: no privilege boundary, flat memory, and sequential tool calls. Capability-based security with an Isolated Tool Runtime, DAG-based parallel tool execution, MemGPT-style tiered memory (working, recall, archival, observational), and progressive tool disclosure via fuzzy search.
3 tachi-agent Local-First Runtime TypeScript ~300 A pluggable ReAct hub that fuses persistent memory (dokoro) with a multi-model council (TachiBot) over MCP — and defaults to 100% local inference via Qwen2.5 on Ollama. The agent loop lives in the client, never inside an MCP server, keeping the hub small and embeddable. Memory is a "bookend" by default (recall before, log after) but can be moved into the loop with a single flag.
4 unimcp MCP Aggregator TypeScript ~400 Solves MCP config duplication and tool visibility without server-side profiles. One config, one shared daemon, per-client tool control via environment variables. Client-side UNIMCP_INCLUDE / UNIMCP_EXCLUDE means each client declares its own tool surface. Auto-spawns on first use, shuts down after 30s idle, and hot-reloads on config change — no orphaned Docker containers or zombie stdio processes.
5 DeepMCPAgent MCP Agent Framework Python ~800 A plug-and-play framework for dynamic MCP tool discovery that works with any LLM and connects to multiple MCP servers simultaneously. Tools are typed through a three-stage pipeline (JSON-Schema → Pydantic → LangChain BaseTool), making MCP tools first-class citizens in LangGraph. Supports HTTP/SSE servers with headers and auth, and ships with a CLI for interactive agent chat.
6 PicoClaw (Sipeed) Ultra-Light Runtime Go ~2k An independent Go implementation of a personal AI assistant that runs in less than 10 MB of RAM and can be deployed on a Raspberry Pi. 95% of the core code was agent-generated through a self-bootstrapping process, making it both a functional runtime and a case study in autonomous AI development.
7 NanoBot Educational Runtime Python ~1.5k Roughly 4,000 lines of pure Python implementing over 90% of OpenClaw's core capabilities — WebUI, tools, long-term memory, MCP integrations, model routing, and multi-agent delegation. The clearest codebase for learning how an agent harness actually works. pip install ready, no compilation, and a structure that maps directly to the conceptual components of an agent loop.
8 ZeroClaw Zero-Overhead Runtime Rust ~1.2k A single-binary Rust runtime with a memory footprint of ~7.8 MB. Four traits (Provider / Channel / Memory / Tool) define the entire architecture. Trait-driven design means every component is swappable. Connects to ~20 LLM providers, runs on Discord/Telegram/Slack, and has near-instant cold starts. The compilation requirement is the only barrier.
9 magg MCP Aggregator Python ~600 A meta-MCP server that aggregates multiple MCP services into a single unified endpoint — essentially giving AI assistants the power to extend their own capabilities on demand. Minimal dependencies, lightweight proxy architecture, and a clean separation between configured MCP sources and the unified interface presented to clients. A simpler alternative to unimcp for Python-native stacks.
10 HarnessRouter Multi-Harness Router Python ~700 The self-hosted, Apache-2.0 unified interface for running multiple agent harnesses behind one API. Instead of committing to one harness, you route tasks to whichever harness fits. Solves the harness lock-in problem directly. Run Claude Code, Codex, and Gemini CLI side-by-side with a single control plane. Still early, but the architectural premise is sound.
11 emergent-wisdom/orchestrator Multi-Agent Engine Python ~300 A lightweight multi-agent orchestration engine built around an "Inbox Zero" architecture and stigmergic collaboration — agents coordinate through a message bus rather than a central planner. Three orchestration modes (queue-based, two-phase, four-phase pipelines) with native MCP support. The stigmergic model — where agents leave traces in the environment that influence others — is a genuinely different coordination primitive.
12 stagechain Pipeline Engine Python ~200 A generic engine for chaining and routing tasks between AI agents — or any callable steps at all. It was built to demonstrate orchestration and pipeline-design patterns as reusable infrastructure, not as a wrapper around a specific application. Framework-agnostic by design. You can mix LLM calls, deterministic functions, and agent invocations in the same pipeline. Useful for teams that want orchestration without adopting a full agent framework.
13 brAIn Event-Triggered Orchestration Python ~150 A node-based agent framework where orchestration is event-triggered, not scheduler-driven. If your agent's cadence is "every Monday morning summarize X," brAIn treats the schedule as a primitive rather than bolting it on. Reactive many-to-many nodes with criticality-based preemption. The event-first model is a better fit for ambient agents that respond to signals rather than run on timers.
14 niflheim-x Lightweight Orchestration Python ~250 A composable agent orchestration framework positioned as "the fast alternative to LangChain." Installable via pip3 install niflheim-x. Focused on reducing the overhead of LangChain-style abstractions while preserving composability. A good fit for teams that found LangChain too heavy but still want structured orchestration.
15 JL-Engine-Local Dynamic Assembly Engine — ~200 A dynamic agent-assembly engine that builds and runs AI agents entirely in RAM, wiring up tools and behavior on the fly. No persistent agent definitions — the agent is assembled per task. The in-RAM assembly model makes it possible to create ephemeral agents with precise tool surfaces for each invocation, reducing context bloat and eliminating state leakage between tasks.
16 Honcho Self-Hosted Memory Layer Python ~400 A self-hosted deployment of Honcho, the cross-session memory layer that powers Hermes Agent's L4 memory. Runs the full stack (API, Deriver, PostgreSQL, Redis) on your own machine with no code changes to Hermes. Data sovereignty without sacrificing memory quality. Routes LLM calls through any OpenAI-compatible provider, and the tradeoff versus Plastic Labs' Neuromancer models is well-documented (slightly less precise deductive reasoning, same observation extraction).
17 runtimenoteslabs/memory-layer Outcome-Based Memory Python ~300 Persistent memory for coding agents with outcome-based learning. It doesn't just store what happened — it learns which memories actually help by tracking whether recalled advice led to successful outcomes. The outcome-based learning loop is the differentiator. Most memory layers store and retrieve; this one closes the feedback loop between memory retrieval and task success.
18 nautilus-compass Black-Box Memory TypeScript ~250 The only public memory layer that doesn't burn LLM tokens to extract facts before storing. It treats memory as a black box and detects drift without an extraction step. Eliminates the token cost of memory ingestion entirely. For agents that run frequently and generate large conversation logs, this is a meaningful cost reduction.
19 MemoryLayer Verified Memory — ~200 Local-first AI memory with reasoning and formal verification. It reasons over what it's told, catches contradictions, derives facts you never stated, and checks every claim before your agent repeats it. Drops into Claude Code, Cursor, and Windsurf via one MCP config line. The formal verification layer is unusual in the memory space — most memory layers trust what they store.
20 oh-my-pi / can1357 Coding Agent TypeScript ~2.6k A terminal coding agent with hash-anchored edits, an optimized tool harness, LSP integration, and subagent support. It doesn't try to be a platform — it tries to be a very good terminal agent. Hash-anchored edits solve a real problem in coding agents: diffs that break when line numbers shift. The LSP integration gives the agent semantic understanding of the codebase rather than just text.

What These Projects Have in Common

The strongest hidden gems share three design instincts:

Small, auditable cores. SimonAgent, NanoBot, and ZeroClaw all prioritize a codebase small enough to read in an afternoon. This isn't just aesthetic — it means the security boundary is inspectable, the failure modes are enumerable, and the agent loop isn't buried under framework abstractions.

One sharp architectural bet. DragonScale bets on capability-based secu

  • permalink
  • fedilink
  • source
    •