# OpenClaw **Aliases:** Warelay, Clawdbot, Moltbot **Entity class:** Open-source personal AI agent platform **Creator:** Peter Steinberger **Steward:** OpenClaw Foundation **License:** MIT **Maturity:** Evolving ## Definition **OpenClaw** is open-source infrastructure for running a persistent AI assistant on an operator's own computer or shared gateway and reaching it through messaging channels and native applications. It connects pluggable models and agent harnesses to state, memory, credentials, tools, skills, plugins, channels, and device-local actions. Its importance to the personal-agent category is architectural and cultural. OpenClaw made the agent visible as a maintained software actor with a runtime, memory, tools, communication channels, and continuing responsibilities—not merely as a chat response produced by one model. ## Name and Project Lineage The project's own vision document records the direct name lineage: **Warelay → Clawdbot → Moltbot → OpenClaw** It began in late 2025 as Peter Steinberger's personal weekend project for an assistant that could perform real tasks on a real computer. In 2026 stewardship moved to the nonprofit OpenClaw Foundation while the code remained open and model-neutral. This is the documented project lineage. Its broader architectural inheritance is conceptual rather than a claim of direct source descent: conversational assistants supplied the interface, LLM tool use supplied action selection, messaging bots supplied ubiquitous channels, local-first software supplied operator ownership, and agent harnesses supplied the loop among planning, tools, state, and environmental feedback. ## Architecture OpenClaw centers a **Gateway** that acts as the control plane for sessions, tools, events, and channel connections. The control UI, command-line interface, terminal interface, messaging services, and companion applications connect through that gateway. Models and harnesses are swappable plugins; tools, skills, channels, and plugins extend the action surface. The architecture separates several objects: - **runtime:** a persistent gateway rather than a one-response process; - **channels:** WhatsApp, Telegram, Slack, Discord, iMessage, Teams, and other communication surfaces; - **models and harnesses:** pluggable reasoning backends rather than one mandatory model; - **state and memory:** operator-controlled persistent context; - **tools, skills, and plugins:** executable capabilities and reusable procedures; - **nodes and companion apps:** device-local voice, screen, camera, canvas, and action surfaces; - **policy and security:** pairing, sandboxing, exposure controls, secrets handling, and operator-defined boundaries. OpenClaw's core distinction is ownership: state, memory, and credentials can remain on hardware controlled by the operator. That design contrasts with managed personal-agent services whose runtime and memory live in a provider's cloud. ## Influence and Descendants OpenClaw's influence appears in both infrastructure and product design: - **NemoClaw** packages OpenClaw with NVIDIA's open models and secure runtime for private, always-on deployments. - **Microsoft Scout** was presented by the OpenClaw Foundation as an OpenClaw-powered or OpenClaw-inspired enterprise personal assistant, with Microsoft contributing upstream. - The **OpenClaw Foundation** is developing a neutral standards and ecosystem role around agent identity, profiles, evaluation, skills, security, and enterprise deployment. - OpenClaw normalized the expectation that a personal agent can be persistent, reachable through ordinary channels, extensible through skills, and separable from any single model vendor. ## Competitors and Adjacent Systems - **[[wiki/Grok Bot|Grok Bot]]** competes through a managed service organized around persistent named bots, team coordination, hosted computers, skills, and routines. - **[[wiki/Meta Muse|Muse]]** competes through mass-market onboarding, a dedicated managed secure VM, connected consumer applications, approvals, payments, and distribution through Meta products. - **[[wiki/Facebook M|Facebook M]]** is a historical precursor to the intent-to-outcome abstraction but relied on a human execution layer rather than an open persistent-agent runtime. These products overlap without being interchangeable. OpenClaw emphasizes operator ownership, extensibility, model choice, and local control; Grok Bot emphasizes a hosted roster of durable workers; Muse emphasizes consumer accessibility and managed trust infrastructure. ## Relationships - **category:** [[wiki/Personal AI Agent|Personal AI Agent]] and [[wiki/Agentic AI|Agentic AI]]. - **architecture:** [[wiki/Agentic Memory|Agentic Memory]], [[wiki/Agentic Workflow|Agentic Workflow]], [[wiki/Model Context Protocol|Model Context Protocol]], and [[wiki/Approval Architecture|Approval Architecture]]. - **comparators:** [[wiki/Grok Bot|Grok Bot]] and [[wiki/Meta Muse|Muse]]. - **human consequence:** [[wiki/Agentic Empowerment|Agentic Empowerment]] and [[wiki/Agent-Management Literacy|Agent-Management Literacy]]. ## Sources / Provenance - [OpenClaw — GitHub repository and architecture](https://github.com/openclaw/openclaw) - [OpenClaw — Vision and documented name lineage](https://github.com/openclaw/openclaw/blob/main/VISION.md) - [OpenClaw — Why OpenClaw](https://docs.openclaw.ai/start/why-openclaw) - [OpenClaw Foundation — Introducing the OpenClaw Foundation](https://openclaw.ai/blog/introducing-openclaw-foundation) - [[wired/Wired About the Age of Agentic Empowerment|Wired About the Age of Agentic Empowerment]] **As of:** 2026-09-28