Common questions.

A bot is one bridge for one platform. openclawOS is the bridge for every platform — same identity, same memory, same agent — across all your messengers. Plus it's an agent, not just a chatbot: Pi can use tools (shell, browser, file system), summon sub-agents, and remember things across sessions.

openclawOS is the Neul Labs distribution. Same kernel-based architecture, same channels, same MIT license. Differences: prebuilt Docker images, signed binaries on Homebrew, a dedicated Discord, and managed hosting on the way. Upstream improvements rebase into our fork weekly.

Yes. The Gateway runs on hardware you control. Conversations, memory and credentials live in local SQLite. There is no openclawOS telemetry. The only data that leaves your network goes to your chosen LLM provider when you ask Pi a question — same as if you were using that provider directly.

Anthropic (Claude — recommended), OpenAI (GPT-4, Codex, Responses), Google (Gemini), GLM, MiniMax, and anything OpenAI-compatible (Ollama, vLLM, Together, Groq, Replicate). You can mix providers across different bindings.

Yes — point openclawOS at an Ollama or vLLM endpoint. Tool use quality varies by model; for full Pi functionality we recommend a Claude-grade model. For chat-only, Llama 3.1 70B works well.

Install the binary (one command), open the Control UI on localhost:7777, paste an LLM API key, pair a channel by following its 30-second wizard, and start messaging. Most users are talking to Pi in under 5 minutes total.

For most channels, no. WhatsApp, Telegram, Discord, iMessage and Signal all work over outbound connections. Slack and Microsoft Teams need a publicly reachable webhook URL — easiest with Cloudflare Tunnel or Tailscale Funnel.

Yes — the Gateway runs on a Pi 4 (4GB+) comfortably. iMessage requires macOS, so for iMessage support you'll want a Mac mini instead.

openclawOS uses Baileys (WhatsApp Web protocol) which respects standard message-rate limits. For very-high volume you'd graduate to the official WhatsApp Business API (which openclawOS also supports).

Implement the tiny Agent interface in TypeScript and drop it under ~/.openclaw/agents/. Hot-reload picks it up. Or fork Pi and patch its skills. The docs at docs.openclaw.ai have a full tutorial.

Not yet, but it's in the works. Join the waitlist at /pricing — we'll run dedicated VMs in your region with backups, SSO and an SLA.