Type /create in @OpenClawBoxBot. Pick a plan. Pay with Telegram Stars. Your secure, isolated OpenClaw instance deploys on Kubernetes in under 60 seconds, with built-in model access and cloud Chrome for real web tasks.
From zero to your own AI assistant in four steps. No server management. No Docker. No nginx. Just Telegram.
Find @OpenClawBoxBot in Telegram or click the link. No signup, no email, no account creation.
One command. That's it. The bot shows you three plans — BYOK, Pro, or Max — each with different AI models and resources.
Tap to pay with Telegram Stars. Native Telegram payment — no credit card, no Stripe, no redirect. Telegram takes 0% commission.
Your isolated OpenClaw instance spins up on Kubernetes in under 60 seconds. Get your personal URL and start chatting with your AI across Telegram, WhatsApp, Discord, Slack, and more.
Self-hosting · 4+ hours
$ apt install docker nginx certbot
$ docker-compose up -d
$ certbot --nginx -d my.domain.com
$ ufw allow 443 && ufw enable
$ vim /etc/nginx/sites-available/...
$ systemctl enable --now openclaw
# configure DNS, SSL, updates, backups...
OpenClawBox · under 60 seconds
We strip away all the complexity of hosting an AI agent. OpenClaw Box isn't just a server—it's a fully managed, isolated environment designed specifically for the OpenClaw AI assistant, complete with a powerful Cloud Chrome integration.
No server setup, no Docker, no nginx configs. Type /create in @OpenClawBoxBot, pick a plan, and your isolated AI instance is live on Kubernetes in under 60 seconds.
No credit cards, no Stripe, no redirect pages. Pay with Telegram Stars right inside the chat - Telegram takes 0% commission. Crypto payments coming soon.
Your instance runs on managed Kubernetes in the cloud. No downtime when your laptop sleeps. Always-on, auto-scaled, monitored, and backed up. We handle the infra so you don't have to.
You get Google Chrome in the cloud for your tenant. Log in safely to websites, keep sessions alive between tasks, and let OpenClaw control the same browser when work goes beyond chat.
We provide access to GPT-5.2, GPT-5, Grok-4, DeepSeek R1, and 10+ models on both Pro and Max plans. No need to create OpenAI or Anthropic accounts, no API key setup. Just start chatting.
Log in to websites inside a remote Chrome session and let OpenClaw operate that same browser. Most competitors stop at basic hosting and leave browser work unsolved.
We are OpenClaw. Same-day support for new releases, direct access to core developers, and auto-updates on day one.
Connect Model Context Protocol servers on any subscription. Bring your own tools and data sources without upgrades.
Deploy once, chat everywhere - Telegram, WhatsApp, Discord, Slack, iMessage, Signal, Web, and Matrix.
Not just a single assistant - deploy a swarm of preconfigured AI agents that coordinate over Slack. Each agent has a focused role, clean context, and specialized tools.
Launch notes, architecture updates, cloud browser milestones, and the weekly changes shipping into OpenClaw Box.
A simple macOS guide to configuring self-hosted OpenClaw browser profiles. Learn the difference between `openclaw` and `user`, how to switch the default profile, how to configure Brave correctly, and how to troubleshoot Chrome MCP attach issues.
Mar 27, 2026 • 7 min read
Learn how to connect Claude Code to your active browser session with Chrome DevTools MCP so it can reuse tabs, cookies, and logged-in web context.
Prefer a feed reader? Subscribe to RSS.
We publish launch notes, browser milestones, and weekly release notes.
Deploy a swarm of preconfigured OpenClaw agents that coordinate autonomously over Slack. Each agent has a focused role, clean context window, and specialized toolset — no context flooding, no single-agent bottleneck.
Morning team! Goal for today: fix the webhook latency issue.
I've flagged 5 related Zendesk tickets and linked them to the GitHub issue.
On it. I'll optimize the database query and open a PR.
I'll monitor the staging deployment once the PR is merged.
Support Engineer
Sentry monitoring, customer email triage, issue escalation, status updates.
Software Engineer
Bug diagnosis & fixes, feature implementation, code review, test writing.
Release Engineer
Release coordination, CI/CD pipelines, safe rollouts, incident response.
Product Manager
Backlog prioritization, PRD writing, roadmap planning, stakeholder updates.
Marketing Manager
Release announcements, social media, blog content, public communications.
Any role you need
Define custom roles with your own knowledge base, tools, and personality. The team adapts to your operations.
A single agent handling support, engineering, deployments, and marketing fills its context window with irrelevant information. An agentic team gives each agent a clean, focused context — better decisions, parallel execution, and domain expertise.
Each agent has a clean, focused context window. No cross-domain noise degrading model performance.
Multiple agents work simultaneously. One patches code while another prepares the deploy pipeline.
Every decision and action is visible in Slack. Intervene anytime, override any agent.
Define custom roles, knowledge bases, tools, and personality. The team adapts to your operations.
Powered by OpenClaw agents coordinating over Slack, GitHub, and Discord.
Enterprise-grade infrastructure on Azure Kubernetes Service. Per-tenant namespace isolation, GenAI proxy with spend controls, automatic TLS, and zero shared state between instances.
Telegram API
getUpdates / webhook
@OpenClawBoxBot
grammY · Node.js 22 · TypeScript
Tenant Gateway
{id}.oclawbox.com/api/chat
PostgreSQL
Users, subscriptions, payments, tenant state
Telegram Stars API
createInvoiceLink · pre_checkout_query · XTR
LiteLLM Proxy
genai.oclawbox.com · per-tenant keys · budget caps
OpenAI
GPT-5.2, GPT-5, o3
xAI
Grok-4, Grok-4 Fast
DeepSeek
R1, V3.2
BYOK
Any OpenAI-compatible
Azure Kubernetes Service (AKS)
a1b2c3.oclawbox.com
OpenClaw Gateway Pod
PersistentVolume · NetworkPolicy
Pro plan
d4e5f6.oclawbox.com
OpenClaw Gateway Pod
PersistentVolume · NetworkPolicy
Max plan
g7h8i9.oclawbox.com
OpenClaw Gateway Pod
PersistentVolume · NetworkPolicy
BYOK plan
Telegram updates arrive via long-poll, the bot resolves the tenant namespace and forwards to the OpenClaw Gateway endpoint. Control plane stores only routing metadata and subscription state.
All model requests route through LiteLLM with per-tenant virtual keys, spend limits, and provider failover. BYOK users bypass the proxy and connect directly to their provider.
Each instance runs in a dedicated Kubernetes namespace with its own PersistentVolume, ResourceQuota (CPU/RAM/storage), and NetworkPolicy denying cross-namespace traffic.
Chat history, RAG documents, and user files stay inside the tenant PersistentVolume. The control plane retains only: Telegram user ID, plan tier, billing records, and routing config.
Every instance runs in its own Kubernetes namespace with dedicated CPU, RAM, and storage limits. Network policies prevent cross-tenant traffic. Your data never touches another user's instance.
Traefik provisions Let's Encrypt certificates automatically. Cloudflare handles DNS. Every tenant gets a subdomain (id.oclawbox.com) with HTTPS on day one.
We roll out new OpenClaw releases with canary deployments — zero downtime. Daily backups protect your data. Rollback if anything breaks.
Pay with Telegram Stars. No credit card. No Stripe. No hidden fees. Telegram takes 0% commission.
1 Telegram Star ≈ $0.013 USD. Buy Stars via Apple/Google IAP or @PremiumBot.
~$10/month
For developers and personal use
AI Models
Bring Your Own Keys — connect any AI provider (OpenAI, Anthropic, Google, Azure, local LLMs)
~$25/month
For power users who want hosted models
AI Models
GPT-5.2, GPT-5, GPT-5.1, GPT-5-mini, GPT-4.1, GPT-oss-120b, Grok-4, Grok-4 Fast Reasoning, Grok-4 Fast, DeepSeek R1, DeepSeek V3.2
+ $25/mo AI credits
~$100/month
For teams and heavy workloads
AI Models
All Pro models — GPT-5.2, GPT-5, GPT-5.1, GPT-5-mini, GPT-4.1, GPT-oss-120b, Grok-4, Grok-4 Fast Reasoning, Grok-4 Fast, DeepSeek R1, DeepSeek V3.2
+ $100/mo AI credits
Selected plan
Host type
Hire
/create and enter the promo code.Frontier models from OpenAI, xAI, and DeepSeek — all available on Pro and Max plans via LiteLLM. The only difference is credit amount ($25/mo vs $100/mo). Or bring your own API keys on the BYOK plan.
Pro and Max plans include hosted models via genai.oclawbox.com (LiteLLM proxy). Each user gets a personal API key with monthly budget tracking. Use /usage to check your spend.
Everything is controlled via chat commands. No dashboard, no web portal — just Telegram. Here's the full command reference.
/createKey CommandCreate a new OpenClaw instance. Pick a plan, pay with Stars, and your AI deploys in 60 seconds. The only command you need to get started.
/startSame as /create. Also handles Telegram deep links for referrals.
/plansView all available plans (BYOK, Pro, Max) with pricing, models, and resource details.
/statusCheck your subscription plan, expiration date, instance URL, hosted models, and monthly credits.
/usageView your AI model usage and remaining credits (Pro & Max plans). See per-model spend breakdown.
/restartRestart your OpenClaw gateway. Useful if something gets stuck.
/logsFetch the last 20 lines of your instance's pod logs for debugging.
/upgradeUpgrade to a higher plan. Sends a new Stars invoice for the upgrade.
/downgradeDowngrade to a lower plan. Takes effect at the next billing cycle.
/cancelCancel your subscription. Instance stays active until expiration, then gets suspended.
/deletePermanently delete your instance. Removes all data, revokes API access, ends subscription.
/referralGet your referral link. Earn bonus Stars when friends subscribe.
/helpShow the full command list with descriptions.
/supportGet support contact info — Telegram channel, email, and payment support.
/termsView Terms of Service.
Beyond commands, just send any message to the bot — text, voice, photos, documents, videos, stickers, GIFs — and it gets forwarded directly to your OpenClaw instance. Chat with your AI naturally, right in Telegram.