Maritimemaritime

The cloud built for
AI agents.

Start for freeRead the docs
The Maritime agent dashboard: live status, CPU / memory / network telemetry, the serving panel, and streaming logs

Each agent on its own micro-VM: live status, logs, and a shell, sleeping when idle. From $1 a month.

Builders from these organizations run agents on Maritime

GoogleSamsungL'OréalTikTokPhilipsDP WorldLiberty MutualTata Consultancy ServicesWixByteDanceMITHarvardCoat of arms of AndorraGovern
d'AndorraGovTech SingaporeEmeritus

Built to be the best place to run an agent.

Deploy in seconds, watch it run, SSH in when you need to. Starting at a dollar a month.

One command
to a running agent.

Maritime detects your framework, builds an isolated micro-VM, and puts your agent live. No infra to wire up.

Cheap: from $1 a month

Run one agent or hundreds. Each VM is sized to its work and sleeps when idle, so you never pay for a server sitting warm.

Convenient: one CLI, one SDK

A real CLI, a typed SDK, a dashboard, and a library of templates to start from. No glue code.

API endpoints

Any framework, any model

Run OpenClaw, Hermes, ZeroClaw, or your own stack. Bring whatever model and tools you already use.

Fast: live in seconds

One command from your terminal to a running agent. It sleeps when idle and wakes the moment traffic arrives.

Encrypted by default

Every Template comes with encrypted secret storage for API keys and credentials. Injected at runtime, never exposed.

OPENAI_API_KEY********************
REDIS_URL********************
DATABASE_URL********************
ANTHROPIC_API_KEY********************
GOOGLE_AUTH_SECRET********************

Reliable: survives your launch

Capacity is added automatically as traffic grows. A launch-day spike is handled before you notice it.

Three steps to a running agent.

Start from a template or bring your own code, deploy with one command, and let Maritime add capacity as traffic grows. You don't touch the infra.

Pick a template
HBS Case PrepOpenClaw
Lead ResearchHermes
Personal AdminOpenClaw
Clone a working agent from the library.
Customize with AI
Make it answer in Spanish
Editing system prompt...
Translated 12 cases. Done.
Add a LinkedIn lookup before each case
Added tool: linkedin_search
OAuth wired to your account ✓
Tell it what to change, in plain English.
Deploy & scale
One command. It sleeps when idle and wakes on traffic.

Start from a template, not from scratch.

Templates are ready-made agents. Clone one, change a few lines, deploy. Each is a real, working agent: a starting point, not a toy.

Sales agent

Qualifies leads, books meetings, follows up. The classic first product to sell.

Support agent

Answers from your docs, escalates when it should, learns your tone.

Research swarm

Many agents searching, reading, and reporting back in parallel.

Agent simulation

Spin up hundreds of agents to model a market, a crowd, or a protocol.

Coding agent

A long-running agent that writes, tests, and ships code on a schedule.

Data agent

Watches a source, transforms, and pushes results downstream, around the clock.

Built one yourself? Publish it. Maritime handles the hosting, scaling, and sharing.

One platform, from your first agent to your whole company.

The platform is the same whether you run one agent or a few thousand.

Developers

Ship one agent for a dollar. Build it with whatever framework and tools you already use. Deploy it, and it's yours.

  • ›SSH into the machine, read every log
  • ›Any LLM, any framework
  • ›One agent or a whole swarm
1import { OpenClaw } from 'maritime'
2
3const agent = new OpenClaw({
4 version: '3.11',
5 model: 'gpt-5.4',
6 sandbox: false,
7 persistence: true,
8})
9
10await agent.deploy()
Deploy your first agent

Startups

Sell agents to your customers. Build one, add a few lines of code, and Maritime runs a copy for every customer who signs up.

  • ›One isolated micro-VM per customer
  • ›Optional LLM proxy in front
  • ›A thousand overnight sign-ups, handled
See the startup setup

Enterprises

Run Maritime on-prem. Install it on hardware you already own and host thousands of isolated agents.

  • ›Your data stays on your infrastructure
  • ›Massive density on existing servers
  • ›Auto-scaling and isolation inside your network
MemberRoleAgents
Talk to us

Every agent gets its own machine.

Each agent you deploy gets a dedicated micro-VM on a bare metal host. Maritime packs the hosts tightly, so a machine of your own starts at $1/month.

Isolation

A hard machine boundary around each agent. No shared kernel, no shared memory.

Ownership

You get root on a machine you can SSH into. Copy everything out whenever you want.

Efficiency

Agents sleep when idle and wake on demand. Packing sleeping VMs tightly is what keeps a machine of your own so cheap.

How an agent costs
a dollar a month.

We size each micro-VM to its work and sleep it when idle. That's the difference between a dollar and a monthly server bill.

MaritimeMaritime
AWS Lambda
AWS EC2
Google Cloud
DigitalOcean
Railway
Heroku
Fly.io
Render
Monthly cost
$1/mo flat
$18+/mo
API Gateway + NAT extras
$8-15/mo
Pays even when idle
$6-49/mo
Always-on hits $49+
$16/mo
You manage the server
$5-10/mo
Usage-based, ephemeral
$7-25/mo
Add-ons push costs up
$6+/mo
IPv4 + volumes extra
$7+/mo
Database costs extra
Setup time
Minutes
Hours
Hours
Hours
Minutes
Minutes
Minutes
Medium
Minutes
No cold starts
None
Yes
None
Yes
None
None
Yes
None
None
Stateful containers
Yes
No
Yes
No
Yes
No
No
Yes
No
Sleep / wake
Built-in
No
No
No
No
No
No
No
No
Agent-first design
Yes
No
No
No
No
No
No
No
No

Prices are for an always-on AI agent (~1 vCPU, 1 GB RAM) as of March 2026: approximate monthly on-demand rates in US regions; they may vary. AWS Lambda figures include API Gateway and NAT Gateway for VPC-attached workloads. Maritime's $1 flat rate comes from sleep/wake: agents use compute only while handling requests.

No lock-in. Not the model, not the framework, not the machine.

my-agent: env

Open or closed, hosted or your own weights. Switch whenever you want.

Deploy your first
agent for $1.

Install the CLI, pick a template, and ship. Scale when you're ready and only pay for what you run.

Start for $1Talk to us about on-prem