ez agent

set up an AI agent for your own tasks

runs fully automatically in the background

automates the tasks you choose

uses the tools and permissions you provide

executes the workflows you configure

uses the tools and permissions you provide

self-refines its routines

operates asynchronously

run it on your own hardware or a cloud host · model and hosting costs vary

note that this is mostly untested, highly experimental and possibly dangerous WIP

ez.lol is not responsible for any of your or your bot's actions

expert mode

step 1: choose your bot engine

Hermes Agent

OpenClaw: an assistant built on Node.js with messaging and tool integrations

ZeroClaw: a lightweight agent implementation a lightweight agent implementation written in Rust

Hermes Agent: an agent by Nous Research with reusable skills, memory and task automation

step 2: choose your deployment pathway

look for a current model with strong tool use. compare recent evaluations and new models, check your engine’s compatibility, and choose what fits your hardware or API budget.

cloud API (easiest)
use a hosted model without running inference yourself. create an API key at OpenRouter or another supported provider; check its current models, limits and pricing.

local LLM
100% private and free, but trickier and takes more time to set up. requires decent hardware (adequate RAM/GPU) or a high-performance cloud server run inference on bare metal. requires dedicated VRAM or a high-performance cloud compute instance

expert free API hunting
explore free-for.dev/ai and the free LLM API repo

notebook (temporary testing)
start your own workspace on Colab, Kaggle or another notebook host. choose a model that fits its hardware; availability, session limits and persistence vary.

API local notebook

step 3: choose your deployment OS

choose the system that will run the bot. a hosted API needs no local GPU; local models need enough memory. for a remote host, check current plans at Oracle Cloud, AWS or Google Cloud.

step 4: install and connect

follow the current installation instructions for your system.

openclaw setup

install the selected engine using its official guide, then complete its onboarding.
choose your provider, paste its API key and select a supported model.
set up Ollama, LM Studio or vLLM. choose a model that fits your hardware and configure its endpoint in the bot.
create a notebook workspace and follow the vLLM quickstart to serve a compatible model that fits the available memory. configure the bot with the server’s model name, endpoint and credentials. keep both running in the same workspace; sessions may stop or reset.
open the bot’s local dashboard and send a test message before adding tools or background tasks.

step 5: make it useful (optional)

connect a supported messaging app if you want to use the bot away from its dashboard.
add the tools and permissions needed for your tasks, then test each workflow.
enable the engine’s background service on a persistent host for scheduled work. a temporary notebook is not a reliable always-on host.
review activity, model usage and costs in the bot and provider dashboards.