Pricing

Pricing

One agent per subscription. Pick a billing cadence that suits you — the product is identical on both.

AgentRoost Starter

For solo builders shipping their first agent

$10 in monthly LLM credits

≈ 3,400 AI replies · 200 deep agent tasks

↻ Fresh credits every month · top-ups never expire

See live prices for all 350+ models →
$19.99/mo
  • 1 vCPU · 2 GB RAM · 20 GB disk
  • Pick from 350+ LLM models
  • Top-ups never expire
  • Priority support
  • Cancel anytime

AgentRoost Pro

For production workloads at scale

$80 in monthly LLM credits

≈ 27,000 AI replies · 1,700 deep agent tasks

↻ Fresh credits every month · top-ups never expire

See live prices for all 350+ models →
$99.99/mo
  • 4 vCPU · 8 GB RAM · 80 GB disk
  • Pick from 350+ LLM models
  • Top-ups never expire
  • Priority support
  • Cancel anytime
Raw workspace

Raw workspace

$9.99/month

Billed every month. Cancel anytime from your account.

  • One dedicated agent workspace
  • Hermes and OpenClaw runtimes preinstalled — pick one, configure with your own LLM key
  • Persistent NVMe SSD per workspace
  • DDR5 ECC memory, Intel Core Ultra cores
  • Public subdomain on agentroost.app
  • Three-tier egress firewall
  • Monitored 24/7, always-on infrastructure
Get started
14-day refund windowCancel anytimePolar is merchant of record

Everything that ships with a managed agent

Beyond the per-tier credits — what we actually handle so you don't.

Hermes or OpenClaw, pre-wired

Pick your framework in the wizard. Either way, memory, scheduling, and tool-use hooks come already configured — no YAML to write, no Docker image to build, no shim to maintain.

Auto-provisioned Telegram bot

We mint the bot on our manager account at checkout. No BotFather conversation, no token handling, no webhook plumbing.

Customizable personality

Describe your agent's voice + behavior in plain language. Edits hot-reload — no restart, no redeploy.

350+ LLM models, one click to switch

Swap from Haiku to Opus to Llama mid-conversation. Per-call cost shown in the picker (platform fee included).

Scheduled tasks built in

Cron jobs for daily summaries, reminders, scraping runs — your agent picks them up automatically. No external scheduler.

Memory that persists

Conversation context, learned facts, file uploads survive every restart. Workspace state lives on a dedicated NVMe volume.

From checkout to first message in 3 steps

No infrastructure to plan, no keys to rotate, no integrations to wire up.

1

Pick a tier

Choose Base, Pro, or Scale. Polar handles checkout — your card statement reads "Polar".

2

Name + personalize

In the wizard, give your agent a name and describe how it should behave. Pick a starting LLM model.

3

Connect with Telegram

One click opens our manager bot, /start your new agent, and you're live. Message it like any other contact.

≈ 2 minutes end-to-end · No DevOps required

FAQ

Common questions before you sign up

The five things customers ask before checking out.

What happens if I burn through my credits mid-month?

Your agent keeps running. Add a one-time top-up pack ($5, $10, $25, or $50) from the agent dashboard — top-ups never expire, so leftover balance rolls forward into future cycles. The included monthly allowance resets on your billing anniversary.

Can I switch between Base, Pro, and Scale?

Each managed agent is tied to one subscription tier. You can't change tier in place on an existing agent — instead, cancel the current sub (workspace is preserved for 7 days) or open a new subscription at the tier you want and run a second agent alongside the first.

Which model should I start with?

Claude Haiku 4.5 is the best balance of cost and quality for everyday conversational agents — that's why our tier credit estimates are based on it. Switch to Opus 4.6 for harder reasoning tasks and pay for it from the same credit pool. The model picker shows the per-Mtok cost (platform fee included) up-front.

Are any models off-limits to managed agents?

No — all 350+ models from the OpenRouter catalog are available to every managed tier. Free models stay free (they don't draw from your credit balance). The catalog refreshes every 6 hours, so anything OpenRouter adds, we expose.

Hermes, OpenClaw, or n8n — which framework should I pick?

You choose in the launch wizard, and every framework comes at the same tier price. Want a personal assistant in chat? Hermes is our proven default and OpenClaw is the popular open-source agent — both managed Telegram bots. Want visual workflow automation? n8n gives you your own instance — your login, your data — on the same dedicated workspace. Same credits, same always-on runtime either way.

What if I want to leave? Can I export my data?

Yes. From settings, you can export everything we hold about your account as JSON (chat history, agent configs, billing records) before deleting the account. After deletion, workspaces and disks are torn down per the Refund Policy. We don't silently archive anything.

Compare Plans

FeatureAgentRoost StarterAgentRoost PlusAgentRoost Pro
Compute1 vCPU · 2 GB RAM · 20 GB disk2 vCPU · 4 GB RAM · 40 GB disk4 vCPU · 8 GB RAM · 80 GB disk
Monthly LLM credits$10/mo$35/mo$80/mo
AI replies (Haiku 4.5)~3,400~12,000~27,000
Deep agent tasks (Opus 4.6)~200~750~1,700
Pick from 350+ LLM models
Top-ups never expire
Priority support
Cancel anytime

Which runtime is right for you?

Every tier runs any of the three. Same price, same credits — pick the one that matches how you want to build. You can change your mind later.

HermesProven default
What it is
Our longest-running agent runtime — persistent memory, connector hooks, and scheduling, fully managed on your own Telegram bot.
What you control
Personality, model choice, and connectors — all from chat. We handle the bot, the keys, and the hardware.
Pick it when
You want a capable assistant that just works and you are not sure which to choose. The safe default.
OpenClawOpen-source
What it is
The viral open-source personal agent, running 24/7 on dedicated hardware and auto-provisioned on your own Telegram bot.
What you control
Naming, starting model, and personality — set up in the wizard, no Docker or YAML.
Pick it when
You like the open-source project and want it hosted and metered for you, with zero DevOps.
n8nYour own instance
What it is
The popular workflow-automation tool, running as YOUR OWN instance on a dedicated always-on workspace we provision for you.
What you control
Your login, your data, your workflows — credentials, schedules, and webhooks, all in your own editor on your subdomain.
Pick it when
You want visual, multi-step automations and a web editor instead of a chat bot.

14-day refund window

Both plans are eligible for a full refund within the first 14 days of the most recent charge — no questions asked, request it from [email protected]. After 14 days, payments for the current billing period (monthly or annual) are non-refundable. You can still cancel anytime to stop the next renewal.

How the annual plan is charged

Annual subscribers are charged the full $99.90 at checkout. The subscription runs for 12 months and renews automatically at the same price. You can cancel auto-renewal at any time from your account; cancellation stops the next renewal and does not refund the current term.

Prices in USD. Polar.sh handles payment and is the merchant of record — your card statement reads "Polar".