Solo $29, Pro $99, or Team $199. Every plan includes all daemons and tools.
Stripe or crypto. Server provisions with the full osModa stack in minutes.
Chat with your workers in natural language or SSH in with full root access.
osModa plans: Dedicated Servers from $29/mo
Every osModa plans plan gives your AI workers a dedicated Hetzner server with the full self-healing stack: NixOS, 10 daemons, watchdog auto-restart, atomic rollbacks, tamper-proof audit, and all 92 tools. No per-token charges. No credit systems. No feature gates between tiers. The only difference is hardware.
In 2026, most agent hosting platforms charge $199–$200/month for shared sandboxes with credit-based limits. osModa plans starts at $29/month and gives you a dedicated server with root SSH access. You bring your own LLM API keys, so you control both infrastructure costs and inference costs independently. Or self-host the entire stack for free — it is open source under the Apache-2.0 license.
TL;DR
- • Three plans: Solo ($29), Pro ($99), Team ($199) -- the only difference between tiers is server hardware.
- • Every plan includes the full stack: 10 daemons, watchdog, atomic rollbacks, audit ledger, and all 92 tools.
- • No per-token fees, no credit systems, no usage caps -- flat monthly rate, cancel anytime.
- • Up to $1,644/yr cheaper than Perplexity Computer ($200/mo) or Manus AI ($199/mo) with dedicated (not shared) infrastructure.
- • 100% open source (Apache-2.0) -- self-host free on your own hardware or use managed hosting at spawn.os.moda.
Choose Your Plan
All plans include the complete osModa stack. Pick the hardware that fits your workload.
Team
$199/mo
8 vCPU · 16 GB RAM · 160 GB NVMe
5–10 agents
Multi-agent workflows, heavy loads.
Get StartedScale
$299/mo
16 vCPU · 32 GB RAM · 320 GB NVMe
10–20+ agents
Full AI workforce, unlimited potential.
Get StartedAll prices in USD. Billed monthly. Cancel anytime. No setup fees.
What's Included in Every Plan
There are no feature gates between tiers. Every plan ships with the entire osModa stack. The only variable is server hardware.
Full root SSH access
Your server, your rules. Inspect, debug, configure anything.
NixOS + 10 daemons
Complete worker platform with declarative config and zero-GC runtime.
Watchdog auto-restart
6-second median recovery. Crashed workers restart automatically.
Atomic rollbacks
Revert any bad deployment instantly. No partial states.
Tamper-proof audit ledger
SHA-256 hash-chained log of every worker action. Immutable.
92 built-in tools
File ops, HTTP, process management, secrets — all native Rust.
Secrets management
Encrypted injection of API keys and credentials at runtime.
Dedicated Hetzner server
No multi-tenancy. No noisy neighbors. Your own machine.
All frameworks supported
LangGraph, CrewAI, AutoGen, MCP, custom Python/Node.js.
Telegram + multi-channel
Manage workers via Telegram, Discord, Slack, WhatsApp, or SSH.
For a deep dive into each feature, see the AI agent hosting overview and our self-healing servers page.
How AI Agent Hosting Pricing Compares
Mainstream agent platforms charge $199–$200/month for shared infrastructure with credit-based limits and no root access. osModa plans gives you a dedicated server with the full stack for a fraction of the cost.
| osModa plans | Perplexity Computer | Manus AI | |
|---|---|---|---|
| Starting price | $29/mo | $200/mo | $199/mo |
| Infrastructure | Dedicated server | Shared sandbox | Shared sandbox |
| Root SSH access | Yes | No | No |
| Billing model | Flat monthly rate | Subscription | Credits + subscription |
| Self-healing | 6-second watchdog | Platform-managed | Platform-managed |
| Atomic rollbacks | NixOS built-in | No | No |
| Audit ledger | SHA-256 tamper-proof | Limited | Limited |
| Open source | Yes (Apache-2.0) | No | No |
| Annual savings vs. competitors | Up to $1,644/yr | — | — |
For detailed head-to-head breakdowns, see platform comparisons, including osModa vs Perplexity Computer and osModa vs Manus AI.
The Real Cost of Hosting AI Workers
Most teams underestimate the hidden costs of worker infrastructure. Here is what running a production AI worker actually costs across different approaches.
DIY on a VPS
$5–$50/mo
Cheap hardware, but you build everything: process supervision, crash recovery, secrets management, audit logging, security hardening. Weeks of engineering before your first worker runs. And when it crashes at 3am, nobody restarts it.
osModa plans
$29–$199/mo
Dedicated server with the complete worker platform pre-installed. Self-healing, audit, 92 tools, root SSH. Production-ready typically under 20 minutes. The sweet spot between cost and operational maturity.
Shared Platforms
$199–$200/mo
Turnkey but expensive. Shared infrastructure, no root access, credit-based limits, vendor lock-in. No way to inspect what is running on your machine because it is not your machine.
Switching from Perplexity Computer or Manus AI to osModa plans saves up to $1,644 per year — while gaining root access, atomic rollbacks, and tamper-proof auditing that those platforms do not offer.
Self-Host for Free (Apache-2.0)
osModa is fully open source. If you have your own servers, you can install the entire stack — all 10 daemons, 92 tools, NixOS configuration, audit ledger — for free. No license fees. No call-home telemetry. No restrictions.
curl -fsSL https://get.os.moda | shWhat you get
The exact same software that runs on managed servers. Same 10 daemons, same 92 tools, same NixOS atomic rollbacks, same watchdog. No feature is held back for paying customers.
When to go managed
The managed service at spawn.os.moda handles provisioning, hardware management, and Hetzner infrastructure. Choose managed if you want turnkey deployment without managing your own bare metal.
Browse the source at github.com/bolivian-peru/os-moda. Read the deployment guide for step-by-step installation instructions.
Frequently Asked Questions About AI Agent Hosting Pricing
How does billing work?
You are billed monthly on the anniversary of your signup date. Each server is a separate line item. There are no per-token charges, no credit systems, and no usage caps on the hosting itself. You pay a flat rate for your server resources, and everything included in the plan is yours to use without limits. Cancel anytime with no early termination fees.
Can I upgrade or downgrade my plan?
Yes. You can change plans at any time from the spawn.os.moda dashboard. Upgrades take effect immediately — your server is reprovisioned with the new resource allocation, and your data is migrated automatically. Downgrades take effect at the start of your next billing cycle. NixOS atomic switching ensures zero downtime during resource changes.
What is included in every plan?
Every plan includes a dedicated Hetzner server, full root SSH access, the complete NixOS + 10 daemons stack, watchdog auto-restart with 6-second recovery, atomic rollbacks, tamper-proof SHA-256 audit ledger, all 92 built-in tools, and secrets management. There are no feature gates between tiers — the only difference is server hardware resources (vCPU, RAM, NVMe storage).
How do I choose the right plan?
Solo ($29/mo) works well for solo builders running 1–3 lightweight workers, hobby projects, and development environments. Pro ($99/mo) suits teams running 3–8 concurrent workers in production. Team ($199/mo) is for AI agencies and high-volume operations running 8–20+ workers, or workloads that need significant memory and storage.
Is there a free tier or trial?
There is no free tier for the managed hosting service. However, osModa is 100% open source under the Apache-2.0 license. You can self-host it on any Linux server for free — download from github.com/bolivian-peru/os-moda and install with a single curl command. The managed service at spawn.os.moda adds the convenience of automated provisioning, dedicated Hetzner infrastructure, and turnkey setup.
Why is osModa so much cheaper than Perplexity or Manus?
Perplexity Computer ($200/mo) and Manus AI ($199/mo) bundle AI inference costs into their pricing and run on shared infrastructure. osModa is hosting-only — you bring your own API keys for LLM providers (OpenAI, Anthropic, etc.), so you only pay for the compute you actually use. Plus, each Spawn Worker runs on a dedicated Hetzner machine, not a shared sandbox, giving you better performance isolation at a fraction of the cost.
What happens if I exceed my server resources?
osModa runs on dedicated hardware, so there is no surprise throttling or overage billing. If your workers consistently need more CPU or RAM than your current plan provides, you will see performance degradation (slower execution, potential OOM kills). The watchdog daemon will keep restarting processes, and the audit ledger will log resource pressure events. That is your signal to upgrade to the next tier.
Do I pay for AI model API calls separately?
Yes. osModa is infrastructure hosting, not an AI inference provider. You bring your own API keys for whatever LLM providers your workers use (OpenAI, Anthropic, Google, open-source models, etc.). Your workers call those APIs directly from your dedicated server. This means you have full control over model selection, rate limits, and costs — and you are never locked into a single provider.
Can I run multiple workers on one server?
Yes. The osModa worker supervisor can manage multiple worker processes on a single server. Each process gets its own watchdog monitoring, audit trail, and secrets scope. How many workers you can run depends on their resource requirements and your plan. A Worker server can handle 1–3 lightweight workers, while a Fleet server can comfortably run 8–20+ concurrent workers depending on workload.
Is there a discount for annual billing?
Not currently. All plans are month-to-month with no commitment. We keep pricing simple: one flat monthly rate, cancel anytime. This means you can experiment with different tiers and scale up or down as your needs change without worrying about locked-in contracts.
Start Hosting Your AI Workers Today
Dedicated servers. Self-healing runtime. Tamper-proof audit. All 92 tools. Full root SSH. From $29/month — or self-host for free.
Last updated: May 2026