Quick Start

1Install the CLI

Install the Asterism CLI globally using npm, yarn, or pnpm:

Or with yarn: yarn global add @asterlabs/cli

2Find Skills

Search for skills that match your workflow. No account needed.

This surfaces skills like scripting-tiktok-videos. Or browse skills on the skills page.

3Install an Agent

Install a complete, ready-made agent — every skill in it is security-screened. No account needed.

(pack: is the install prefix for a whole agent's skill set.) Browse the full catalog at /agents, or install a single skill instead: aster install scripting-tiktok-videos. Everything lands in the detected agent's skills directory — e.g. ./.claude/skills/ for Claude Code (or ~/.claude/skills/ with --global).

4Deploy a whole agent definition

aster install copies skill files. aster deploy materializes a whole agent definition — persona, skills, permissions and MCP wiring — into a harness's own native config:

That writes AGENTS.md, the skills into .agents/skills/, and a .codex/config.toml in which the agent's declared permissions become Codex's own primitives — an fs:read agent becomes sandbox_mode = "read-only". Swap --harness claude-code and the same model becomes a deny list in .claude/settings.json.

Claude Code and the OpenAI Codex CLI are the two harnesses verified end to end. Adapters for openclaw, hermes and eve exist but have never been run against a live install — treat them as beta.

5Connect to Your AI Assistant

To let your assistant pull skills from the registry at runtime, add the MCP server to .mcp.json in your project root (the same file aster deploy writes):

// .mcp.json
{
  "mcpServers": {
    "aster": {
      "command": "npx",
      "args": ["-y", "@asterlabs/mcp"],
      "env": {
        "ASTER_API_KEY": "your-api-key"
      }
    }
  }
}

Get your API key from the dashboard.

6Authenticate (Optional)

Log in when you want to publish skills, leave reviews, or access private skills:

This will open a browser window for authentication. You can use GitHub or Google.