How it works

How AgentPod works

Your agent runs in its own isolated, backed-up container in the cloud — driven from Telegram, Slack, or your browser, with a real Chrome and your Google Workspace built in.

The journey

From a voice note to a finished file

You talk to your AgentPod in Telegram. It does the work in a real browser inside your own container — signed in as you — and sends the result back to the same chat.

  1. 1

    You send a voice message

    Hold the mic in Telegram and say what you need — no app to install, no console to open. Voice, text, photos, and files all reach your agent the same way.

  2. 2

    Your agent opens its cloud browser

    Every AgentPod runs a real Google Chrome inside your own isolated container. Your agent drives that browser: it opens pages, searches, reads, clicks — the open web, not a scraping API.

  3. 3

    It works inside your Google Workspace

    Sync the cookies from the browser you are already signed into and the cloud Chrome is signed in as you. Your agent reads the unread thread in Gmail, checks Calendar, opens Drive — as you, with your access.

  4. 4

    It writes the result into a Google Doc

    Findings do not stop at a chat message. Your agent writes them into a real document you can open, edit, and share with your team.

  5. 5

    The finished file comes back to Telegram

    The document lands in the same chat you started in. One thread, start to finish — the work happened somewhere else, but you never had to go there.

End to end

The architecture, end to end

From your laptop's identity, to the channel you ask from, to the container that does the work — here is the full path a request takes.

❯ How it actually works

Ask from the web app, Telegram or Slack. Your agent — OpenClaw or Hermes, your choice — drives a cloud Chrome that's already signed in as you, then gets real work done across Gmail, Calendar, Docs, Drive and more.

your isolated LXD sandboxLocal laptopyour browser, signed inSynchronize cookies to allow access to the logged inservice on the cloud browser for AI agentWeb appTelegramSlackOpenClawswap engines anytimehermes-aiswap engines anytimeChrome Browsersigned in as youGoogle Workspacesigned in as youComposio5 apps at launchCalendarGmailDocsDriveExplore hundreds of apps in the Composio catalogue ↗Salesforce · Jira · Stripe · Linear · Airtable · HubSpot · Shopify · Supabase · Asana · Attio · Apollo · Amplitude · Algolia · Ashby · 1Password · Alpha Vantage · ActiveCampaign · Affinity · Ably · Ayrshare · Salesforce · Jira · Stripe · Linear · Airtable · HubSpot · Shopify · Supabase · Asana · Attio · Apollo · Amplitude · Algolia · Ashby · 1Password · Alpha Vantage · ActiveCampaign · Affinity · Ably · Ayrshare ·
Security & isolation

Your own container. Nobody else's data.

AgentPod runs OpenClaw, Hermes-AI, and Paperclip inside isolated LXD Linux containers — full OS-level isolation with automatic backups, so security is not something you have to configure yourself.

Isolated LXD container

Every agent runs in its own dedicated LXD Linux container — full OS-level isolation, not a shared process or a chroot. Your container is yours alone.

Zero shared state

No cross-tenant traffic, no shared database rows, no shared filesystem. Each container has its own storage, network policy, and resource limits — nothing leaks between users.

Automatic backups

Your workspace, files, browser state, and configuration are backed up automatically, so a restart, an upgrade, or a failure never costs you your work.

Per-tenant TLS & encryption

Every container gets its own subdomain with automatic HTTPS. Traffic and data at rest are encrypted, so nothing between you and your agent is ever sent in the clear.

Ready to try it yourself?

Start a free 3-day trial in Telegram, or open the web console — no credit card required.

Try in TelegramTry in a browser
See why AgentPod beats a plain VPS