What is OpenClaw

Self-hosted, model-agnostic personal AI agent (by Peter Steinberger) that gets full computer access — browser, filesystem, shell — and is reachable from 29+ chat platforms, positioned against SaaS agent walled gardens.

Updated July 9, 2026 · §202607081949-9 ·

OpenClaw is an open-source personal AI agent created by Peter Steinberger that runs locally on your own machine (macOS, Linux, Windows) rather than in a vendor’s cloud. It’s pitched as “a 24/7 assistant with access to its own computer” — the emphasis is on doing real things (sending emails, filling forms, running scripts), not just chatting.


Core design choices

ChoiceWhat it buys you
Runs locallyState (memory, credentials, task history) stays on your machine, not a SaaS vendor’s servers
Model-agnostic gatewaySwap Claude, GPT, or a local model behind the same agent without rewriting anything
Multi-channelOne agent, reachable from 29+ platforms — WhatsApp, Telegram, Discord, Slack, Signal, iMessage
Skills frameworkCommunity-contributed or self-authored plugins via ClawHub; the agent can write its own skills
You (WhatsApp / Slack / Discord / ...)


   OpenClaw gateway  ──▶  Model of your choice (Claude / GPT / local)

          ├── Browser control (fill forms, extract data)
          ├── Filesystem read/write
          ├── Shell / script execution
          └── Cron jobs (proactive, unattended tasks)

What people actually use it for

Reported use cases skew toward personal ops automation: inbox triage, calendar coordination, flight check-ins, expense reporting and tax prep, website building, code review, test automation, and orchestrating multiple agents across a Discord server.

The positioning: self-hosted vs. SaaS agent

OpenClaw’s pitch is explicitly a reaction against cloud agent platforms: “your context and skills live on YOUR computer, not a walled garden.” That’s the same self-hosting argument What is Hermes Agent makes, and both projects compete on the same axis — control and data locality — rather than raw capability.

AxisOpenClawWhat is Hermes Agent
OriginPeter Steinberger (indie)Nous Research (lab, tied to Hermes models)
Memory modelPersistent, skills-framework driven”Holographic” fact memory (SQLite/FTS5) with a closed-loop skill-distillation step
Distribution channels29+ chat platformsTelegram, Discord, Slack, WhatsApp, Signal, CLI
Extension modelClawHub community skills/pluginsAuto-distilled skills from its own task history

Note on claims: OpenClaw’s own marketing cites very large adoption numbers (GitHub stars in the hundreds of thousands) for a project this young. Treat vendor-reported growth figures for any fast-moving open-source AI project as directional, not verified, until you’ve checked the repo yourself.

Why it’s on the backlog: it’s the most “consumer-facing” entry in this list — useful as a reference point for how far a self-hosted agent can reach into personal systems (browser + shell

Local graph

Full graph →