Step 1: Create Your Account
Go to spawn.os.moda and create an account. No credit card required for the initial signup. The dashboard shows your servers, health status, and audit logs.
osModa uses passwordless authentication. You provide your SSH public key during account creation, and that key becomes your access credential for both the dashboard and your servers. One key, no passwords, no MFA token to lose.
Step 2: Pick a Plan
All plans include the full osModa stack. The difference is server resources. Choose based on what your agent needs:
| Plan | RAM | vCPU | Disk | Best For |
|---|---|---|---|---|
| Starter | 4 GB | 2 | 40 GB | API-based agents, lightweight bots |
| Builder | 8 GB | 4 | 80 GB | RAG agents, multi-step workflows |
| Pro | 16 GB | 6 | 160 GB | Multi-agent systems, data pipelines |
| Scale | 32 GB | 8 | 320 GB | Local LLMs, heavy compute workloads |
Plans start at $14.99/month. Every plan includes: self-healing watchdog with sub-6-second recovery, SHA-256 tamper-evident audit logging, encrypted mesh networking (Noise_XX + ML-KEM-768), NixOS atomic rollback, and dedicated Hetzner hardware. No shared tenancy. See the full pricing details.
Step 3: SSH Into Your Server
After provisioning (60-90 seconds), you receive an IP address. SSH in with the key you provided during signup:
ssh root@your-server-ip
You land in a full NixOS environment. The osModa stack is already running:
# Check osModa services systemctl status osmoda-watchdog # Self-healing watchdog systemctl status osmoda-ledger # Audit logging daemon systemctl status osmoda-mesh # Mesh networking # All three should show "active (running)"
You have root access. Install anything you need. The server is yours — osModa manages the infrastructure health layer; you manage your agent.
Step 4: Install Your Framework
Install whatever your agent needs. Here are examples for the most popular frameworks:
LangChain / LangGraph
# Create a virtual environment python3 -m venv /opt/agent/venv source /opt/agent/venv/bin/activate # Install LangChain + LangGraph pip install langchain langgraph langchain-openai # Set your API key echo 'export OPENAI_API_KEY="sk-..."' >> /opt/agent/.env
CrewAI
pip install crewai crewai-tools
OpenAI Agents SDK
pip install openai-agents
Custom Node.js Agent
# Node.js is available via nix nix-env -iA nixpkgs.nodejs_22 # Initialize your project mkdir -p /opt/agent && cd /opt/agent npm init -y npm install @anthropic-ai/sdk express
The frameworks page has detailed setup guides for every major agent framework.
Step 5: Deploy Your Agent
Create a systemd service to run your agent as a managed process. This gives you automatic restart, log management, and integration with the osModa watchdog.
# Create the service file cat > /etc/systemd/system/my-agent.service << 'EOF' [Unit] Description=My AI Agent After=network-online.target Wants=network-online.target [Service] Type=simple User=root WorkingDirectory=/opt/agent EnvironmentFile=/opt/agent/.env ExecStart=/opt/agent/venv/bin/python main.py Restart=on-failure RestartSec=5 MemoryMax=3G [Install] WantedBy=multi-user.target EOF # Enable and start systemctl daemon-reload systemctl enable --now my-agent # Verify it's running systemctl status my-agent
That is it. Your agent is running as a managed process with automatic restart on failure, memory limits, and logging via journald. The osModa watchdog picks up the new service and begins health monitoring automatically.
Alternative: Self-Host with the osModa Installer
If you have your own servers (on-premises, existing cloud instances, or GPU machines), you can install the osModa stack directly:
curl -sSf https://install.os.moda | sh
This installs the 9 Rust daemons that comprise the osModa infrastructure layer: the self-healing watchdog, SHA-256 audit ledger, mesh networking, and supporting services. It works on any NixOS-compatible system.
Self-hosting gives you the same reliability guarantees as the managed service — same watchdog, same audit logging, same mesh networking — on hardware you control. This is the path for teams with GPU servers, specific compliance requirements, or existing infrastructure investments. See the deployment guide for detailed self-hosting instructions.
What osModa Gives You Out of the Box
Every osModa deployment — managed or self-hosted — includes these infrastructure capabilities from first boot. You do not configure them. They work automatically.
Self-Healing Watchdog
Monitors your agent processes with configurable health checks. Detects crashes, hangs, and degraded performance. Automatically restarts unhealthy agents within 6 seconds. If a deployment causes repeated crashes, SafeSwitch triggers a NixOS atomic rollback to the last known-good system state. Your agent recovers without human intervention.
SHA-256 Audit Ledger
Every agent action, tool invocation, and system event is recorded in a hash-chained ledger. Each entry references the hash of the previous entry, making the log tamper-evident by construction. If any entry is modified, the chain breaks and the tampering is detectable. This is your compliance artifact for SOC 2, HIPAA, or any framework requiring audit controls.
Encrypted Mesh Networking
Agent-to-agent communication encrypted with Noise_XX protocol and ML-KEM-768 post-quantum key encapsulation. Your agents talk to each other securely without you configuring TLS certificates, VPNs, or firewall rules. Peer discovery and key exchange happen automatically within the mesh.
NixOS Atomic Rollback
Every system configuration change creates a new NixOS generation. If anything goes wrong — a bad package update, a configuration error, a broken dependency — the system rolls back to the previous generation in seconds. No manual intervention required. This is infrastructure-level undo that works at the NixOS level, not just the application level.
Verify Everything Works
After deploying your agent, run through this checklist:
# 1. Agent is running systemctl status my-agent # Should show "active (running)" # 2. Watchdog is monitoring it journalctl -u osmoda-watchdog --since "5 min ago" | grep my-agent # Should show health check entries # 3. Audit ledger is recording journalctl -u osmoda-ledger --since "5 min ago" # Should show agent action entries # 4. Test crash recovery kill -9 $(pidof python) # Wait 6 seconds, then: systemctl status my-agent # Should show "active (running)" again # 5. Check logs journalctl -u my-agent -f # Should show your agent's output
If all five checks pass, your agent is production-ready. The self-healing infrastructure handles crashes, the audit ledger records actions, and the mesh is ready for multi-agent communication if you add more servers.
Common Agent Deployment Patterns
Here are the three most common ways teams deploy agents on osModa, based on what we see in production:
Single Agent, Single Server
The simplest pattern. One server runs one agent. Ideal for customer support bots, monitoring agents, or any single-purpose agent. Start with the Starter plan and scale up if needed.
Multi-Agent, Single Server
Multiple lightweight agents (API-based, not running local models) on a single Pro or Scale server. Each agent runs as a separate systemd service. Good for teams running 3-5 complementary agents that do not need GPU access.
Multi-Agent, Multi-Server Mesh
Multiple servers, each running specialized agents, all connected via osModa's encrypted mesh network. Agents communicate peer-to-peer without routing through a central server. This is the pattern for complex multi-agent systems where agents need to collaborate across isolated compute environments.
Frequently Asked Questions
How long does it take to spawn an AI agent on osModa?
Typically under 20 minutes from account creation to a running agent. Server provisioning takes approximately 15-20 minutes. After that, you SSH in, install your framework, and start your agent process. The osModa stack (watchdog, audit logging, mesh networking) is pre-configured and active from first boot — there is zero infrastructure setup required on your end.
What frameworks can I run on osModa?
Any framework that runs on Linux. The server is a full NixOS environment with root SSH access. Common frameworks deployed on osModa include LangChain/LangGraph, CrewAI, AutoGen, OpenAI Agents SDK, Haystack, and custom Python or Node.js agent code. You can install packages via nix, pip, npm, or any package manager. If it runs on Linux, it runs on osModa.
What does osModa give me that a regular VPS does not?
Three things. First, self-healing: the watchdog monitors your agent process with configurable health checks and automatically restarts on failure, with sub-6-second recovery. If a deployment causes repeated failures, SafeSwitch triggers a NixOS atomic rollback. Second, audit logging: every agent action and tool invocation is recorded in a SHA-256 hash-chained ledger that is tamper-evident by design. Third, mesh networking: encrypted agent-to-agent communication using Noise_XX + ML-KEM-768 post-quantum cryptography. A regular VPS gives you a blank Linux box. osModa gives you a production-ready agent platform.
What are the osModa pricing plans?
Plans start at $14.99/month for a dedicated Hetzner server with 4 GB RAM. All plans include the full osModa stack — self-healing watchdog, audit ledger, mesh networking, and NixOS-based infrastructure. The difference between plans is server resources (RAM, CPU, disk). There are no per-agent fees, no token surcharges, and no hidden costs. You pay for infrastructure, not for agent count.
Can I self-host osModa instead of using the managed service?
Yes. osModa provides a curl one-liner that installs the full stack on any NixOS-compatible server. This gives you the same self-healing watchdog, audit logging, and mesh networking capabilities on your own hardware. Self-hosting is ideal for teams with existing infrastructure, specific compliance requirements, or GPU servers that are not available through the managed service. The managed service at spawn.os.moda handles provisioning and maintenance for you.
Do I get root access to the server?
Yes. Full root SSH access. You can install any software, configure any service, and modify any system setting. osModa runs as a service layer on top of NixOS — it does not restrict your access to the underlying operating system. The server is yours to use however you need. The only thing osModa manages automatically is the infrastructure health layer (watchdog, audit logging, mesh).
What happens if my agent crashes at 3 AM?
The osModa watchdog detects the failure through health check validation and automatically restarts your agent within 6 seconds. If the crash was caused by a bad deployment (the agent keeps crashing in a loop), SafeSwitch triggers a NixOS atomic rollback to the last known-good system state. All crash events are logged in the tamper-evident audit ledger. You can review exactly what happened, when, and what action the watchdog took — without needing to be awake at 3 AM.
Can I run multiple agents on a single osModa server?
Yes. The server is a full Linux environment — you can run as many processes as the hardware supports. Each agent should be managed by a separate systemd service for independent process supervision. The osModa watchdog monitors all registered services. For resource-intensive agents (those running local LLMs or processing large datasets), we recommend one agent per server to avoid resource contention. For lighter API-based agents, a single server can comfortably host 3-5 agents.