Skip to content
This repository was archived by the owner on Jun 23, 2026. It is now read-only.

Spectral-Finance/agentsea

Repository files navigation

AgentSea

Launch any AI coding agent on any cloud — with one command.

curl -fsSL https://agentsea.thegrid.ai/cli/install.sh | bash
agentsea claude digitalocean

That's it: AgentSea provisions a fresh, sandboxed environment in your own cloud account, installs the agent, wires up inference through The Grid API, and drops you into an interactive session.

What is AgentSea?

AgentSea is a command-line launcher for AI coding agents. Instead of manually spinning up a server, installing an agent, and configuring API keys every time, you run a single command and AgentSea handles provisioning, setup, and authentication for you.

Agents run on your infrastructure — your laptop or a VM in your own cloud account — and reach LLMs through The Grid, an OpenAI-compatible inference API. You bring your cloud account and your Grid key; AgentSea does the wiring.

What it does

  • One-command launchagentsea <agent> <cloud> provisions a VM, installs the agent, and connects you.
  • Agent-agnostic — OpenClaw, Claude Code, Codex, OpenCode, Hermes, Kilo Code, and more. Switch agents by changing one word.
  • Bring your own cloud — DigitalOcean, Hetzner, AWS Lightsail, GCP, and local. Your account, your keys, your bill, your data.
  • Fully sandboxed — each agent gets an isolated VM and credential boundary, so sessions never cross-talk.
  • Grid-backed inference — all agents route LLM calls through The Grid API, so budgets, keys, and usage stay on-platform.
  • Run it locally — no cloud account required; agentsea <agent> local installs the agent directly on your machine.

Who is it for

  • Developers who want to spin up a disposable AI coding agent without babysitting servers or copying API keys around.
  • Teams who need each agent run isolated in its own VM and credential scope.
  • Anyone on The Grid who wants to point an agent at Grid-backed inference in seconds.

Quickstart

1. Install the CLI (macOS, Linux, or WSL — needs bash, curl, ssh, jq):

curl -fsSL https://agentsea.thegrid.ai/cli/install.sh | bash

2. Authenticate with The Grid:

agentsea auth login

This runs a browser/device login and saves a consumption key. (You can also set THEGRID_API_KEY manually — create one at https://app.thegrid.ai.)

3. Launch an agent:

# On your own machine — no cloud account needed
agentsea claude local

# On a cloud VM (uses your provider token, e.g. DIGITALOCEAN_ACCESS_TOKEN)
agentsea claude digitalocean

Useful commands:

agentsea                 # interactive agent + cloud picker
agentsea agents          # list available agents
agentsea clouds          # list supported clouds
agentsea list            # browse and rerun past launches
agentsea status          # show live state of your servers
agentsea delete          # tear down a server
agentsea --help          # full reference

Cloud provider tokens use each provider's standard env var (DIGITALOCEAN_ACCESS_TOKEN, HCLOUD_TOKEN, …). Run agentsea <cloud> to see the setup steps for a specific provider.

Repo layout

Path What's inside
manifest.json Public catalogue of agents, clouds, and their availability
sh/ Per-cloud bootstrap scripts (sh/<cloud>/<agent>.sh)
assets/ Agent and cloud icons
packages/cli The agentsea command-line tool
packages/sdk @agentsea/sdk — shared manifest types and loader
packages/ui The marketing site at agentsea.thegrid.ai

Development

npm install

npm run dev          # marketing site → http://localhost:3011
npm run build:cli    # build the CLI bundle
npm run agentsea -- --help

npm run typecheck
npm run lint
npm run build

For local provisioning, copy .env.example.env at the repo root (ignored by git); agentsea loads it automatically when your shell is under this tree, or set AGENTSEA_ROOT to point at it. See todo.md for the public roadmap and CONTRIBUTING.md to get involved.

License

Apache-2.0. See LICENSE.

About

Launch any AI coding agent on any cloud — one command, your cloud account, wired to The Grid API.

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors