Skip to content
ML2
Multi-Step Agent SetupsTool#GitHub repo#Local / BYO model#Open source#Self-hostable#CLI

OpenClaw: a self-hosted personal agent in the chat apps you already use

Added New⧉ Copyable prompt below

Our verdict

A strong choice if you want a personal or small-team assistant that you own: state, memory and credentials stay on your hardware, models are swappable plugins (including local Ollama), and there is no paid tier. The official ClawCast episode 13 on small businesses is a useful starting playbook: connect the tools you already use, watch for work people repeat, and turn it into skills (its examples include proposals, bid calculations, email and calendar briefings, and pulling leads into one place). The main risk is permissions: tools run on the host for the main session unless you set up sandboxing, so read the security guide before you connect other people or expose the Gateway.

Who it's for
Builders and small-business operators who want one assistant across their chat apps, running on a machine they control
Leverage
Repeated business routines turned into skills you own
Effort
Medium
Date added

What it does

OpenClaw runs a local Gateway, the control plane for sessions, tools, events and channel connections, plus a Control UI, CLI and TUI, with companion apps for macOS, iOS, Android, Windows and Linux. It connects to Discord, iMessage, Slack, Teams, Telegram, WhatsApp and more, works with hosted and local model providers (it uses Ollama's native API, not the /v1 endpoint), and is extended with tools, skills and plugins shared on ClawHub. The memory-wiki plugin keeps an Obsidian-friendly knowledge vault, with openclaw wiki obsidian helpers. It is stewarded by the OpenClaw Foundation, a 501(c)(3); by default it only phones home for a daily version check. Release history: the first commit in openclaw/openclaw is dated 24 November 2025, the first GitHub release (v0.1.1) followed on 25 November 2025, and the openclaw npm package was first published on 29 January 2026.

How builders use it

  • Install with the official script, finish onboarding, then open the Control UI with openclaw dashboard.
  • Point it at a local model: ollama pull a model, then openclaw models set ollama/<model>.
  • Connect one channel (e.g. Telegram or Slack) and approve unknown senders with openclaw pairing approve.
  • Follow the ClawCast ep. 13 approach: list the routines your team repeats each week and turn the first one into a skill.
  • Use the memory-wiki plugin if you keep notes in Obsidian, and check it with openclaw wiki obsidian status.

What to expect

A first repeated routine, such as a daily email and calendar briefing, run by an assistant on your own hardware.

Copy this: Install + local model + first check

# macOS / Linux / WSL2
curl -fsSL https://openclaw.ai/install.sh | bash
# (npm route: npm install -g openclaw@latest --allow-scripts=openclaw
#  then: openclaw onboard --install-daemon)

openclaw gateway status
openclaw dashboard

# Local model via Ollama (native API, no /v1):
ollama serve
ollama pull gemma4
export OLLAMA_API_KEY="ollama-local"
openclaw models list --provider ollama
openclaw models set ollama/gemma4

# Approve a new chat sender:
openclaw pairing approve <channel> <code>

# Obsidian vault (memory-wiki plugin):
openclaw wiki obsidian status

Before exposing the Gateway or adding users, read:
https://docs.openclaw.ai/gateway/security
https://docs.openclaw.ai/gateway/sandboxing

Related discoveries

  • ToolCoding & Shipping Agents

    Aider

    Open-source A.I. pair programming in your terminal, with git commits built in.

    Effort: Low#GitHub repo#Local / BYO model

  • ToolCoding & Shipping Agents

    opencode

    An open-source coding agent for the terminal that works with many model providers.

    Effort: Low#GitHub repo#Local / BYO model

  • ToolCreative & Media

    Invoke (InvokeAI)

    An open-source, locally run image studio with a canvas, inpainting and node workflows for open models.

    NewEffort: Medium#Image#GitHub repo

Want more like this?

Practical discoveries, without the noise.

Get the newsletterMore in Multi-Step Agent Setups