Back to directory
3 guides in this path

Getting Started

A six-step path from your first agent prompt to a working local setup. No hype, no vendor lock-in, no skipped fundamentals.

Step 01

Pick one narrow task

Start small. A good first task is: explain a function, generate unit tests, summarize a document, or rewrite one component. The narrower the task, the easier it is to judge whether the agent helped.

Step 02

Try a cloud agent

Use a cloud-hosted agent so you can learn the workflow before worrying about hardware. Match the interface to how you already work.

Browse the Agent Directory
Step 03

Define done and guardrails

Before prompting, write down what 'done' looks like, which files the agent can touch, which commands it can run, and what a human must approve. This habit separates useful workflows from chaotic ones.

Read the safety checklist
Step 04

Compare tools side-by-side

Once you know your task, compare 2–3 agents. Match pricing, runtime, and platform to your stack. Use our comparison pages to keep the evaluation honest.

Compare agents
Step 05

Go local for privacy or cost control

When data cannot leave your machine, move to local models and open-source agents. We publish tested recipes that get you from zero to running without reading ten READMEs.

See deployment recipes
Step 06

Stay current

Agents and models change fast. Bookmark the changelog, tips, tests, and pricing sections — or check back here. We update entries as the landscape shifts.

Browse the changelog

Ready-to-run setups

Jump straight to a working agent

These guides get you from install to first useful conversation in one sitting.

Agent Setup

Getting Started with OpenClaw

Run your own personal AI assistant on your devices and the channels you already use.

Read guide
Agent Setup

Getting Started with Hermes Agent

Install the self-improving, provider-agnostic AI agent that learns from your workflows.

Read guide
Learning

Getting Started with AI Agents

A numbered learning path for choosing, trying, and deploying your first AI agent.

Read guide

Read this before you prompt

Common first mistakes

Too broad a task

Agents struggle with 'fix the codebase.' They excel at 'add input validation to this form.'

No verification

Always run tests, linters, and a human review before applying agent edits.

Ignoring cost

Cloud agents can burn tokens fast. Set a budget or use local models for experimentation.

Trusting blindly

Agents make mistakes. Treat their output as a first draft, not a final answer.