OpenClaw
An AI agent tool built around a skill catalog called ClawHub. Released in November 2025, it features over 137 version iterations and more than 5,400 skills, with support for tool invocation, memory management, and gateway functionality.
Core Facts
Limited by the computing power and memory of consumer-grade devices, the scale of models that can be run locally is typically far smaller than cloud-based flagship models, with gaps in complex reasoning, long-context processing, and other tasks.
A Skill is a modular encapsulation that grants an agent specialized capabilities, containing complete workflows, judgment logic, and contextual knowledge, and can combine multiple tools to accomplish complex goals.
OpenClaw is an open-source AI Agent framework with tool-calling, memory management, and autonomous evolution capabilities.
Timeline (last 90 days)
The turning point in Matt Pocock's attitude toward AI programming came during the release of Opus 4.5 and the formation of the OpenClaw ecosystem, after which he no longer writes code himself.
Products such as Cloud Code, Codex, OpenClaw, Hermos Agent and DeepSync's DSH appeared successively after February 2026, and the industry gradually formed a consensus around Harness Engineering.
During the 2026 Spring Festival, the explosive popularity of OpenClaw (commonly known as "Lobster") brought the agent concept into the mainstream, with both enterprises and individuals trying to use it to improve efficiency.
OpenClaw's documentation warns that smaller or heavily quantized models are more vulnerable to prompt injection attacks and recommends large models for any agent that uses tools.
OpenClaw itself does not produce intelligence; rather, it consumes the inference results of other large models (such as ChatGPT, Claude, Gemini) to perform real tool operations.
OpenClaw's value depends on the reliability of the tools and models it connects to, not on its own level of intelligence.
Builders Creed's video mainly stays at the conceptual level and does not elaborate on OpenClaw's specific installation method, supported model integration protocols, permission isolation mechanisms, or actual task execution success rates.
Understanding OpenClaw as "the operating system layer of AI" is more accurate than understanding it as "yet another AI tool."
Hermes, OpenClaw and Paperclip are all orchestration frameworks or agent runtimes, but each has a different focus.
Crabbox is an open-source tool developed by Peter Steinberger and the OpenClaw team that can push resource-intensive tool calls (such as tests and builds) to remote machines for execution and then sync the state back to the local machine; the agent itself still runs locally.
40 more timeline events
All Facts (20)
Limited by the computing power and memory of consumer-grade devices, the scale of models that can be run locally is typically far smaller than cloud-based flagship models, with gaps in complex reasoning, long-context processing, and other tasks.
90%VerifiedA Skill is a modular encapsulation that grants an agent specialized capabilities, containing complete workflows, judgment logic, and contextual knowledge, and can combine multiple tools to accomplish complex goals.
90%VerifiedOpenClaw is an open-source AI Agent framework with tool-calling, memory management, and autonomous evolution capabilities.
80%VerifiedOpenAI's agents can browse the web, write and execute code, manage files, and complete complex workflows through multi-turn reasoning
75%VerifiedTechnology, tools, and models in the AI field iterate extremely fast, and a guide that has not been updated for half a year may already contain a large amount of outdated content.
65%VerifiedOpenClaw carries security risks: because its built-in tools have extremely high permissions (capable of operating system files, reading browser information, and executing tasks online), any configured API Keys are highly susceptible to leakage if the machine is compromised.
65%VerifiedOpenClaw's architecture is: a person connects to OpenClaw (local computer) via communication software (WhatsApp, Telegram, etc.), which then connects to a language model (cloud/local); OpenClaw itself is an interface program between the person and the language model and does not possess independent intelligence.
65%UnverifiedThe openclaw dashboard command is used to open OpenClaw's built-in Web UI, but it requires the Gateway to be started first in order to load properly.
100%UnverifiedOpenClaw支持SubAgent机制，语言模型可以要求生成子Agent来处理子任务，但程序端禁止子Agent再使用繁殖工具以防止无限递归
95%Unverifiedopenclaw gateway status 命令用于检查OpenClaw Gateway的运行状态，RPC状态显示OK表示服务正常运行
95%UnverifiedOpenClaw使用Compaction机制解决Context Window不够用的问题，将较旧的对话历史发给语言模型做摘要，用摘要替换原文，该过程可递归进行
90%UnverifiedOpenClaw的心跳机制每隔固定时间（如30分钟）自动触发语言模型，让它读取habit.md中的待办事项并执行，配合CronJob排程系统实现定时任务
90%UnverifiedOpenClaw的记忆检索本质上是RAG（检索增强生成），将记忆文件切成Chunk，通过字面匹配和语义Embedding的加权相似度排序，取Top-K结果返回给语言模型
90%UnverifiedOpenClaw中最强大也最危险的工具是Execute，可以执行任何Shell命令，存在语言模型执行破坏性命令（如rm -rf）的风险
90%Unverifiednotebooklm-py可作为Claude Code、Codex、OpenClaw等AI智能体的工具插件
90%UnverifiedClaude Code和OpenClaw都可以将飞书CLI作为外部工具调用，从而获得操作飞书的能力
90%UnverifiedARIS采用纯Markdown技能系统，不依赖任何特定框架，能与Claude Code、Codex、OpenClaw或任意LLM代理配合
90%UnverifiedMoltbot这个名字只存活了三天（2026-01-27至2026-01-30）就被替换为最终的OpenClaw
90%UnverifiedOpenClaw项目经历了六次命名：Warelay（2025-11-24）→ CLAWDIS（2025-12-03）→ CLAWDBOT（2026-01-04）→ Clawdbot（2026-01-10）→ Moltbot（2026-01-27）→ OpenClaw（2026-01-30）
90%UnverifiedOpenClaw项目最初名为Warelay，于2025年11月24日首次提交，定位为基于Twilio的WhatsApp中继CLI工具
90%