Skip to content
Puranode

AI agent hosting

Keep your bots and agents online around the clock

From a Claude-powered ClawdBot to a custom Python agent, Puranode runs it 24/7 with auto-restart, secrets management and persistent storage — no laptop left running.

One-click templates

Launch a known bot, or bring your own

Start from a ready template or deploy any runtime — Node, Python, Go or a Docker image.

AI assistant · most popular

ClawdBot

Claude-powered assistant for Discord & Slack with memory and tools.

Capabilities

Persistent memoryTool useSlash commandsDiscordSlackScheduled jobs
Deploy ClawdBot
Automation

Hermes

Automation & alerting bot — schedules, webhooks and workflows.

Deploy
Node.js

discord.js

Bring your Node.js Discord bot — deploy straight from Git or a zip.

Deploy
Python

discord.py

Python bots with requirements.txt auto-install and a managed venv.

Deploy
Telegram

Telegram bot

Long-polling or webhook Telegram bots, always online.

Deploy
Docker

Custom Docker

Anything containerised — bring your own image and we run it.

Deploy

Always on

Your agent runs 24/7 with automatic restarts if it ever crashes.

Secrets out of your code

Set API keys and tokens as container environment variables in the panel, so they are never committed to your repo or baked into an image.

Persistent storage

Keep memory, databases and files between restarts and deploys.

Deploy from Git

Push to deploy, or upload a zip / Docker image. Logs stream live.

Outbound ready

Full network access for APIs, webhooks and model providers.

Room to grow

Move up a plan as your bot's traffic grows. It is a manual resize we do the same day, during business hours.

Pricing

Pay for the bots you run

Hobby

€4.99/mo
  • 1 bot / agent
  • 0.5 vCPU
  • 512 MB RAM
  • 24/7 uptime
  • Auto-restart on crash
  • Live console logs
Start hosting
Most popular

Standard

€9.99/mo
  • Up to 3 bots
  • 1 vCPU
  • 1 GB RAM
  • Persistent storage
  • Env-var secrets (not in your code)
  • 99.9% uptime target
Start hosting
On request

Scale

€24.99/mo
  • Up to 10 bots
  • 2 vCPU
  • 4 GB RAM
  • Dedicated IP
  • Metrics + alerting
  • Priority deploy queue
Ask us to set it up

Set up by hand once you get in touch.

Need GPU inference or a heavier model runtime? Talk to us.

ClawdBot ready

Have a Claude or Discord bot sitting on your PC? Move it to the cloud.

Migrate my bot

FAQ

AI agent hosting — FAQ

How much does AI agent hosting cost?

AI agent hosting at Puranode starts at €4.99/month for one always-on bot (0.5 vCPU, 512 MB RAM), scaling to €24.99 for up to 10 bots. Flat price, VAT included, cancel anytime.

What can I host — Discord bots, Telegram bots, custom agents?

Yes to all. Run Discord.js/Discord.py bots, Telegram bots, Claude-powered assistants, automation bots, or any containerised app — deploy from Git, a zip or your own Docker image.

Does my bot stay online 24/7?

Yes. Agents run continuously with automatic restart if they ever crash, plus persistent storage. API keys and tokens are set as container environment variables in the panel rather than hardcoded in your code.

Where is it hosted?

On EPYC + NVMe hardware in the EU (Finland), operated by MB VARIANTAI, an EU company in Lithuania. Prices include VAT.

Is 512 MB enough for a Discord bot?

For most bots, yes. A discord.js or discord.py bot handling commands, roles, moderation and API calls sits well inside 512 MB, because it is mostly idle on a socket. What pushes you higher is anything large and resident: a headless browser, image processing, a locally loaded model, or a member cache that grows as your bot joins more servers.

What happens if my bot crashes at three in the morning?

The container restarts it automatically and the console logs record the exit. Worth knowing the limit: a restart repairs a crash, not a bug. A bot failing on a bad token or a missing import will restart into the same failure indefinitely, and it looks like uptime unless you read the console. Restarts also clear anything held only in memory.

Do I need webhooks for a Telegram bot, or is long-polling fine?

Long-polling is fine and needs nothing inbound: your bot calls getUpdates outbound and Telegram holds the request open, keeping undelivered updates for up to 24 hours across restarts. Webhooks need a public HTTPS endpoint on port 443, 80, 88 or 8443 with a trusted certificate, so ask us before you build around one. The two cannot both be active; a leftover webhook makes getUpdates return 409.

Can I run my own Docker image?

Yes, that is what the Custom Docker template is for, and it is the right path for a compiled Go or Rust binary or anything needing system packages such as ffmpeg or Chromium's libraries. Build for linux/amd64 (an Apple Silicon build defaults to arm64 and will not start), run one foreground process as PID 1, and log to stdout and stderr.

Where do my API keys and bot tokens live?

As container environment variables set in the panel, read with process.env or os.environ, so they are never committed to your repository or baked into an image. If a token has already been in git, rotate it in the developer portal rather than editing the file: the old value stays in the history and in every clone and fork of it.

Can I move up to more RAM later?

Yes, by asking. Changing plan RAM is a manual resize we do the same day you ask, during business hours, rather than a self-serve button you can press mid-incident. If you know a launch will need more headroom, tell us in advance. Keep your code in git either way, so a move is a redeploy rather than a recovery.