Documentation

PolymathClaw is a managed hosting platform for OpenClaw, an open-source AI agent platform with 13,700+ skills. We handle the infrastructure — Docker, servers, SSL, and updates — so you can focus on using your AI assistant.

PolymathClaw uses a BYOK (Bring Your Own Key) model. You provide your own API keys from any of our 12 supported providers (Anthropic, OpenAI, Google, xAI, Groq, Mistral, and more), and we host everything for $19/month. This gives you full control over which models you use and direct billing from your provider.

Setup in four steps

1. Sign up & subscribe — create an account and subscribe for $19/month. Cancel anytime from your dashboard.
2. Add an API key — paste at least one API key from any supported provider on your dashboard. We validate the key instantly.
3. Acknowledge API costs — confirm that you understand AI model API costs are billed separately by your provider. We strongly recommend setting a spending limit on your provider's dashboard.
4. Deploy — click "Deploy Instance." Your dedicated OpenClaw instance will be live at yourname.polymathclaw.com in under two minutes.

You need at least one API key to use PolymathClaw. Your keys are used to make AI model requests from your OpenClaw instance — you're billed directly by the provider, not by us.

Supported providers

Adding a key

1. Go to your dashboard and find the API Keys card.
2. Paste your key into the field for the provider (e.g., Anthropic).
3. Click Save. We validate the key with a test request to the provider.
4. If valid, the key is encrypted with AES-256-GCM and stored securely. You'll see a masked hint (e.g., sk-ant-...7x4q).

You can add multiple keys from different providers and switch between their models anytime.

How keys are stored

Your API keys are encrypted with AES-256-GCM at rest. We decrypt keys only to inject them as environment variables into your dedicated container. We never log, share, or access your keys for any other purpose.

Your dashboard is your control center. It shows your instance status and provides management for all aspects of your OpenClaw deployment.

Dashboard cards

• Instance Status — shows whether your instance is running or stopped, with controls to start, stop, and restart. Includes a link to open your WebChat at yourname.polymathclaw.com.
• Model Selection — choose your default AI model from any provider whose key you've added. Changing the model restarts your instance (~10 seconds).
• API Keys — add, update, or remove API keys for any of our 12 supported providers.
• Telegram — connect a Telegram bot to message your assistant from the Telegram app.
• Account — subscription status, billing management via Stripe, and account cancellation.

WebChat & Control UI

Once deployed, your instance has its own URL at yourname.polymathclaw.com. This opens OpenClaw's built-in Control UI where you can:

• Chat with your AI assistant via the WebChat interface
• Configure advanced OpenClaw settings
• Manage skills and channels directly
• View logs and agent activity
• Customize your agent's name, tone, and behavior via the Personality tab

With BYOK, you can use any model your API key supports. There's no plan gating — every model is available to every customer. New models become available as soon as your provider releases them.

Popular models (examples)

Switching models

Use the Model Selection card on your dashboard to change your default model. The dropdown shows models available based on which API keys you've added. Changing the model requires a brief restart (~10 seconds).

API cost estimates

Costs depend on how much you use your agent and which model you choose. AI agents consume more tokens than simple chatbots due to tool calls, system prompts, memory, cron jobs, and sub-agents.

• Budget-friendly — DeepSeek V3.2 or Gemini 2.5 Flash: ~$1–10/month. These models are extremely cheap ($0.15–0.42/1M tokens) and great for everyday tasks.
• Moderate usage — Claude Haiku, GPT-5.2, or Gemini 3.1 Pro: ~$10–40/month. Good balance of quality and cost for most users.
• Power usage — Claude Sonnet, GPT-4o, or o3: ~$30–80/month. Higher quality responses, great for complex tasks.
• Frontier models — Claude Opus 4.6: ~$60–200+/month. Most capable but significantly more expensive ($5/$25 per 1M tokens).

Set a monthly spending limit on your provider account to avoid surprises. Most providers let you set hard caps in their dashboard.

Skills are plugins that extend what your AI assistant can do. OpenClaw has 13,700+ skills available on ClawHub, the community skill marketplace. All skills work out of the box with your managed instance.

What can skills do?

Skills add capabilities like:

• Web browsing — your instance includes a built-in Chromium browser for navigating websites, scraping data, and filling forms.
• Image generation — create images using DALL-E, Stable Diffusion, and other models via skills.
• Code execution — run code in a sandboxed environment.
• Email, calendar, Todoist, Notion, GitHub — connect your productivity tools.
• Cron jobs — schedule recurring tasks (e.g., daily reports, morning briefs).
• Sub-agents — your agent can spawn sub-agents that work in parallel on different parts of a complex task.
• File management — read, write, and organize files in your agent's workspace for persistent data storage.
• Memory — your agent remembers previous conversations and context across sessions.

Installing skills

There are two ways to install skills:

1. From your Control UI — open your instance at yourname.polymathclaw.com, go to the skills section, search, and install.
2. From chat — ask your assistant to install a skill by name (e.g., "install the weather skill"). It will handle the rest.

To browse all available skills, visit clawhub.ai.

Every PolymathClaw instance includes a built-in Chromium browser powered by Playwright. No extra setup, no API key required — it's ready to use the moment your instance deploys.

What it can do

• Browse websites — navigate to any URL, read content, follow links.
• Take screenshots — capture full-page or element-level screenshots.
• Fill out forms — submit contact forms, login to services, complete multi-step flows.
• Extract data — scrape product info, pull tables, gather structured data from pages.
• Monitor pages — check for changes on a page over time (combine with cron jobs for automated monitoring).

Example prompts

• Summarize the top 5 stories on Hacker News right now
• Take a screenshot of example.com and tell me what you see
• Compare the price of AirPods Pro on Amazon vs Best Buy
• Go to my company's status page and tell me if anything is down

How it works

The browser runs inside your container using a real Chromium instance. JavaScript-heavy sites, SPAs, and sites behind client-side rendering all work. No headless browser API key is needed — it's included in your instance.

Your agent can run tasks on a schedule automatically. Just tell it what you want and when — it creates a cron job for you.

How to set up a scheduled task

1. Open your WebChat or Telegram.
2. Describe the task and the schedule in plain language.
3. Your agent confirms the schedule and creates the cron job.
4. The task runs automatically at the scheduled time.

Example prompts

• Every morning at 7am, send me a briefing with top tech news, weather in Calgary, and my calendar for the day
• Every Friday at 5pm, summarize the week's GitHub activity on my repo
• Check job postings on LinkedIn for "senior engineer" roles every hour and alert me about new ones
• Remind me every day at 9am to take my vitamins

Managing scheduled tasks

All management is conversational — just ask your agent:

• Show me all my cron jobs
• Cancel the morning briefing
• Change the weekly GitHub report to Mondays at 9am instead

You can also view and manage cron jobs from the Control UI at yourname.polymathclaw.com.

Your agent remembers context across conversations. Memory persists through restarts, updates, and daily session resets.

What gets remembered

• Facts about you — your name, preferences, timezone, work details you've shared.
• Past conversations — key context from previous chats.
• Instructions — standing instructions you've given (e.g., "always respond in French").
• Task outcomes — results of completed tasks for future reference.

Managing memory

Control what your agent remembers conversationally:

• What do you remember about me?
• Remember that I prefer bullet-point summaries over paragraphs
• Forget everything about my previous project

Session resets vs. memory

Your instance does a daily session reset (default: 4am, or after 2 hours idle) to start a fresh conversation. This clears the active chat thread but does not erase memory. Your agent still knows who you are and what you've told it — it just starts a clean conversation.

Channels are the platforms where your AI assistant receives and responds to messages. OpenClaw natively supports 22+ channels including:

• Web — built-in WebChat at your instance URL. No setup required.
• Telegram — connect a Telegram bot from your dashboard.
• WhatsApp, Slack, Discord, Signal, iMessage, Google Chat, Microsoft Teams, Matrix, IRC — configurable through the Control UI at your instance URL.

We provide dashboard setup for Telegram at launch, with more channels coming to the dashboard soon. You can configure any of the 22+ channels directly from your Control UI.

Setting up Telegram

1. Open @BotFather on Telegram.
2. Send /newbot and follow the prompts to create your bot.
3. BotFather will give you an API token (looks like 123456:ABC-DEF1234...).
4. On your dashboard, find the Telegram card and paste your token.
5. Click "Connect." Your bot will go live in a few seconds.
6. Open your bot in Telegram and send a message — you should get an AI response.

Setting up other channels

For channels beyond Telegram, open your Control UI at yourname.polymathclaw.com and navigate to the channel configuration section. Each channel has its own setup flow — consult the OpenClaw documentation for specific instructions.

Step-by-step guides for popular use cases. Everything is conversational — just tell your agent what you want.

Morning Briefing

Get a personalized daily briefing delivered to your Telegram or WebChat every morning.

1. Connect Telegram (optional) — follow the Telegram setup so your briefing arrives on your phone.
2. Tell your agent what to include — e.g., Every morning at 7am, send me a briefing with top tech news, weather in Calgary, and my calendar summary
3. Done — your agent creates a cron job and delivers the briefing daily.

Tip: Install calendar and weather skills from ClawHub for richer briefings with real calendar data and accurate forecasts.

Research & Summarize

Have your agent research a topic across multiple sources and deliver a structured summary.

1. Describe the topic and depth — e.g., Research the current state of nuclear fusion startups. Check at least 5 sources and give me a summary with key players, funding, and timelines.
2. Your agent browses the web — it visits multiple sites, reads articles, and synthesizes the findings.
3. You get a structured report — with sources cited.

Tip: Use a capable model (Claude Sonnet, GPT-5.2, Gemini Pro) for deep research tasks that require synthesizing many sources.

Code & DevOps

Review PRs, run code, and automate developer workflows.

1. Install the GitHub skill — tell your agent: Install the GitHub skill
2. Authorize — follow the prompts to connect your GitHub account.
3. Use it — e.g., Review the latest PR on my-org/my-repo — summarize the changes and flag any issues

Your agent can also run code in a sandboxed environment — just paste code or ask it to write and execute something.

Email & Communications

Manage email, draft replies, and automate communication tasks.

1. Install an email skill — search ClawHub for Gmail, Outlook, or a generic email skill.
2. Authorize — connect your email account when prompted.
3. Use it — e.g., Summarize my unread emails from today or Draft a reply to Sarah's email about the Q2 budget

Tip: Combine with scheduling to get a daily email summary every morning, or auto-draft replies for review.

One plan: $19/month

PolymathClaw has a single plan at $19/month that includes everything:

• Dedicated OpenClaw instance
• Custom subdomain with SSL
• All 13,700+ skills
• 21 messaging channels
• Browser automation (Chromium)
• Cron jobs, memory, sub-agents
• Automatic updates and restarts
• Data export: download all your agent's data (memory, config, files) as a .tar.gz archive from the Instance panel

What's NOT included: API costs

Managing your subscription

From the Account card on your dashboard, click "Manage Billing" to open the Stripe customer portal. There you can:

• Update your payment method
• View invoices and payment history
• Cancel your subscription

Cancellation

Cancel anytime from your dashboard. Your instance stays active until the end of your current billing period. After that, your instance is shut down and your data is deleted within 30 days.

• Isolated infrastructure — every instance runs in its own dedicated Docker container with isolated resources. No shared environments, no noisy neighbors.
• Encrypted API keys — your keys are encrypted with AES-256-GCM at rest. We decrypt them only to inject as environment variables into your container.
• Encrypted data — all account data is encrypted at rest (AES-256) in Supabase. All data in transit is protected by TLS 1.3.
• Private conversations — your conversations are processed entirely on your isolated instance. We don't store conversation logs on our central systems. Other customers and PolymathClaw staff cannot access your data.
• No tracking — we use only essential cookies (authentication session) and localStorage (theme preference). No tracking cookies, no advertising cookies, no third-party analytics.
• Payment security — payments are processed entirely by Stripe. We store only your Stripe customer ID — never your card number or banking details.

For full details, see our Privacy Policy and Terms of Service.

See our full FAQ on the homepage for more questions. Here are a few common ones: