OpenClaw is the config-first personal AI assistant that runs on any OS and speaks in the channels you already use — WhatsApp, Slack, Telegram, Discord, iMessage, and 20+ more. Point it at ARK and every message, tool call, and session runs inside the EU, with zero retention by default.
OpenClaw is an open-source personal AI assistant that lives on your own hardware and plugs into the channels you already use. No chains, no graphs, no Python glue code — you write a single SOUL.md file that describes who the agent is and what it can do, run one command, and it's online.
The community has built 162+ production-ready agent templates across 19 categories. Each template is a single-file config you can fork, edit, and deploy in minutes.
When you point OpenClaw at ARK, every message your agent sends and every tool call it makes runs against an OpenAI v1-compatible endpoint hosted inside the EU — with zero retention by default and 98.9% fewer tokens per call in stateful mode.
SOUL.md — version-control your agent like any other codeOpenClaw is built for personal, always-on use. That traffic pattern — short bursty calls, session context, multi-channel concurrency — is exactly what ARK's runtime is designed for.
Multi-step reasoning and tool calls re-send the same context every turn. ARK's stateful inference keeps the KV-cache warm per session — 98.9% fewer tokens re-computed per call.
Every channel, every DM, every tool call passes through an EU-hosted inference layer with zero retention by default. Nothing logged, nothing retrained on.
ARK keeps serving even under 99% hardware failure. For an always-on personal agent, that means no dropped messages when a GPU takes a nap.
OpenClaw reads standard OpenAI environment variables. Set your ARK endpoint and key, pick a model, and launch the agent — the same SOUL.md you already have keeps working.
Running ARK Core or Tailored on your own infrastructure? Same config, just swap the hostname for your internal gateway.
Read OpenClaw's setup docs →# Point OpenClaw at the ARK inference gateway: OPENAI_BASE_URL="https://api.ark-labs.cloud/api/v1" OPENAI_API_KEY="ark_sk_live_..." OPENAI_MODEL="llama-3.3-70b-instruct" # Then run your agent: $ openclaw run ./SOUL.md # → Agent online. Listening on WhatsApp, Slack, Telegram. # → Every completion runs in the EU, zero retention.
A single agent in WhatsApp and Slack that drafts replies, schedules calls, and summarizes threads — with your calendar, inbox, and CRM as tools.
A shared agent that answers "what's the status of X," "who owns Y," "book me Z" from team-specific sources — isolated sessions per employee, shared knowledge.
A tenant-isolated agent routing inbound WhatsApp and Telegram messages to workspace-scoped tooling — with full audit trails inside the EU.
Spin up OpenClaw on ARK with free credits — or talk to us about dedicated endpoints and self-hosted deployment.