Configuration

Agents are configured through the CLI and dashboard. Here's what each option means.

Creating an agent

Use maritime create to provision a new agent. You can pass options inline or follow the interactive prompts:

terminal

maritime create my-agent --template openclaw

The New agent dialog in the Maritime dashboard with the template picker and advanced sources — Dashboard equivalent: New agent → pick a template, or deploy from a GitHub repo, Docker image, or ZIP.

Agent options

Option	Flag	Required	Description
name	<name> (positional)	No	Unique agent name (prompted interactively if omitted)
template	-t, --template	No	Template ID: `openclaw`, `maritime`, `hermes`, `zeroclaw`, `openclaw_identity`, `openclaw_browser`, or `hermes_identity`. See `maritime templates`
repo	-r, --repo	No	Deploy from a GitHub repo instead of a template (must contain a Dockerfile). Combine with `-b, --branch`
public	--public	No	Serve a public, no-login web URL for this agent (web apps). Use `--port` to set the port your app listens on (default 8080)
env	-e, --env	No	Initial environment variables as KEY=value (repeatable)

Environment variables

Environment variables are managed through the CLI or the dashboard. Variables are marked as secret by default.

terminal

# Set a secret environment variable
maritime env set my-agent OPENAI_API_KEY=sk-...

# Set a non-secret variable
maritime env set my-agent LOG_LEVEL=debug --no-secret

# List all variables
maritime env list my-agent

# Remove a variable
maritime env remove my-agent LOG_LEVEL

Sets STRIPE_API_KEY as an encrypted secret and LOG_LEVEL as plain text, lists both (the secret stays masked), pulls them into a local .env file, then removes the secret.

Triggers

Triggers (cron, webhook, Telegram, Discord, email) are configured through the dashboard. Triggers wake sleeping agents automatically when fired. List an agent's triggers from the terminal with maritime triggers <agent>.

Resource limits

Every agent gets the same top-tier allocation automatically; there are no plans or tiers to pick. Agents that need more (for example browser agents) are sized up by the platform on their own.

For hand-tuning a single agent, the CLI supports per-agent overrides (--ram, --cpus, --disk, --idle); see the CLI reference.