AgentSea | Launch AI agents wired to The Grid
Launch a supported AI agent locally or on a fresh cloud VM with one command, pre-wired to The Grid. Install the agentsea CLI, add your key, and connect.
AgentSea is a command-line launcher for popular open-source AI agents. One command installs the agent and its dependencies, authenticates with your Grid key, wires the OpenAI-compatible endpoints and model routing, and drops you into the running agent. You can run it on your own machine or on a fresh cloud VM that AgentSea provisions for you.
If you would rather configure an agent by hand, the per-tool guides in Integrations still apply. AgentSea is the fast path to the same result.
Supported agents
Live today: OpenClaw, Hermes Agent, Kilo Code, Claude Code, OpenCode.
Coming soon: Codex CLI, Cursor CLI, Pi, T3 Code, Junie.
Each agent is verified against The Grid on the target environment before launch. For agent-specific configuration detail, see its dedicated guide, for example OpenClaw or Hermes Agent.
Prerequisites
macOS, Linux, or WSL with
bash,curl,ssh, andjqavailable.A Grid account at app.thegrid.ai with a usable consumption balance. The Grid is prepaid; zero credits returns 402.
A Grid consumption API key from Settings → API Keys. Use a consumption key, not a trading key. Trading keys do not authorize inference.
For cloud launches: an account with a supported provider (DigitalOcean, Hetzner, AWS, GCP) and its credentials available locally.
Install
Run the installer in a terminal:
curl -fsSL https://agentsea.thegrid.ai/install.sh | bashWhen it finishes, the agentsea command is on your path. The first launch prompts for your Grid API key and saves it under ~/.config/agentsea/.
Launch an agent
No cloud account needed. AgentSea installs the agent and its dependencies directly on your computer.
AgentSea installs the agent, prompts for your Grid key on first run, sets the environment variables, Grid API endpoints, and model routing, then launches the agent in your terminal with full TTY support.
On each cloud launch, AgentSea provisions a fresh VM in your cloud account and installs the agent via cloud-init. No Terraform or YAML configs to write.
The flow: provision the VM, install the agent and dependencies, inject your Grid key and cloud credentials, configure the OpenAI-compatible endpoints and routing, then open an SSH session so you can drive the agent from your terminal. The VM inherits THEGRID_API_KEY and base URLs targeting the Grid API.
For the exact command flags, provider selection, and per-agent options, see the CLI Reference at agentsea.thegrid.ai.
Manage deployments
Lists your deployments. For agents that ship a web UI, such as Hermes Agent and OpenClaw, use Open Dashboard from the listing.
Model routing
AgentSea wires each agent to The Grid's instruments automatically, so you do not pick a vendor model. If you want to pin or change the instrument an agent uses, follow that agent's dedicated integration guide and the current instruments list.
Verification
Send a prompt in the launched agent. A normal completion confirms end-to-end routing through The Grid. If the call fails, check the troubleshooting items below.
Troubleshooting
401 Unauthorized. The saved key is a Grid trading key, not a consumption key. Re-run authentication and paste a consumption key from Settings → API Keys.
402 Payment Required. Your credits are at zero. Top up at app.thegrid.ai.
agentsea: command not found. The installer did not finish, or your shell has not reloaded. Confirm bash, curl, ssh, and jq are installed, re-run the install command, then open a new terminal.
Cloud provisioning fails. AgentSea could not reach your cloud provider. Confirm the provider is supported (DigitalOcean, Hetzner, AWS, GCP) and that its credentials are available in your local environment.
Last updated
Was this helpful?