Quick Start

Get from zero to a running agent task in under 5 minutes. This guide assumes you have already installed Testy Claw and set your API key.

Step 1 — Install

npm install -g testy-claw

Step 2 — Wake up the agent

Run Testy-Claw wakeup to boot the agent, verify your API key, and confirm all tools are loaded. You'll see a ready prompt when everything checks out.

Testy-Claw wakeup

  ✦ Testy Claw is awake
  ◎ Loading tools     fs · shell
  ◎ Model             gpt-4o  (openai)
  ◎ Config            claw.config.ts
  ✓ API key valid
  ✓ Ready — type your task below

Step 3 — Init your project

Navigate to any project directory and run init. Testy Claw will detect your stack and write a claw.config.ts with sensible defaults.

cd my-project
testy-claw init

  ✓ Detected: Node.js / TypeScript
  ✓ Created claw.config.ts
  ✓ Ready

Step 4 — Review the config

Open claw.config.ts. The defaults work for most projects, but you can customise the model, tools, and limits here.

import { defineConfig } from "testy-claw";

export default defineConfig({
  model: "gpt-4o",          // language model to use
  tools: ["fs", "shell"],   // built-in tools to enable
  maxSteps: 20,             // max agent steps per run
  timeout: 30_000,          // ms before a step times out
});

Step 4 — Run your first task

Pass a plain-English task to testy-claw run. The agent will plan the steps, execute them, and print a summary.

testy-claw run "add a README.md with project name, install instructions, and a usage example"

  ◎ Planning 3 steps...
  ✓ Read package.json to get project name
  ✓ Inferred install command from package.json
  ✓ Wrote README.md (42 lines)
  ✓ Done in 2.1s

Step 5 — Inspect what happened

Every run is saved to .claw/runs/. Use replay to step through the execution log.

testy-claw replay

  Run: 2024-01-15T10:32:11Z
  ─────────────────────────
  Step 1  read_file("package.json")       → { name: "my-project" }
  Step 2  read_file("package.json")       → { scripts: { start: "node index.js" } }
  Step 3  write_file("README.md", ...)    → 42 lines written

What to explore next

• Core Concepts — understand how the agent plans and executes
• CLI Reference — every flag and option for each command
• Built-in Tools — what fs, shell, http, and git can do
• Writing Custom Tools — extend the agent with your own functions
• CI/CD Integration — run agent tasks in your pipeline