How it works
Two layers: your laptop (install the CLI once), then your environment (local machine or cloud VM). For step-by-step deploy commands, pick an agent on the homepage.
One-time install
Paste into a terminal (macOS, Linux, or WSL). When it finishes, the agentsea command should work. Prerequisites: bash, curl, ssh, jq.
curl -fsSL https://agentsea.staging.thegrid.ai/install.sh | bashLocal machine
No cloud account needed. The CLI installs the agent and dependencies directly on your computer.
- 1
Install
Installs the agent and dependencies on this machine. No cloud account needed.
- 2
Authenticate
Prompts for your The Grid API key and saves it under ~/.config/agentsea/ when you confirm.
- 3
Configure
Wires environment variables, Grid API endpoints, and model routing.
- 4
Connect
Launches the agent in your terminal with full TTY support.
Cloud environments
On each launch, the CLI provisions a fresh VM in your cloud account (DigitalOcean, Hetzner, AWS, GCP, …) and installs the agent via cloud-init.
- 1
Provision
Spins up a fresh VM in your cloud account. No Terraform or YAML configs.
- 2
Install
Cloud-init installs the agent and dependencies on the new server.
- 3
Authenticate
Injects your The Grid API key and cloud credentials into the VM.
- 4
Configure
Sets environment, OpenAI-compatible endpoints, and model routing to The Grid.
- 5
Connect
Opens an SSH session. Drive the interactive agent from your terminal.
Bootstrap scripts are served from the CDN at https://agentsea.staging.thegrid.ai/<cloud>/<agent>.sh. The VM inherits THEGRID_API_KEY and OpenAI-compatible base URLs targeting the Grid API.
After deploy
Each agent is verified against The Grid on the target environment. See Grid integration docs for agent-specific setup (OpenClaw, Hermes, Claude Code, Cursor IDE, …).
agentsea ls— list deployments; use Open Dashboard for Hermes / OpenClaw UIsagentsea <agent> <cloud>— provision a fresh environment with Grid credentials wiredTHEGRID_API_KEY— consumption key from app.thegrid.ai