AI Stack Library
How to build an AI operating system, one layer at a time
Each guide answers one question and ends with its sources. Each one also has a plain-text copy your AI agent can read: add .md to the end of the address.
Start here
AI operating system for business
An AI operating system for business is the set of models, agents, memory, rules and schedules that does real work for a company on a machine the company controls.
Agent vs harness
An agent is the job an AI model does when it plans its own steps and calls tools. The harness is the software around the model that runs that job.
Always-on local AI operating system
An always-on local AI operating system is a dedicated machine your business owns that runs its AI agents day and night. Why every business needs one.
AI stack glossary
Plain definitions of the words used to build an AI operating system, from agent and harness to MCP, RAG and model routing, each linked to its full guide.
Rent or own your AI stack
Rent or own your AI stack, layer by layer: own the always-on machine, memory, rules, skills and pipelines; rent frontier models, email, hosting and payments.
Memory
How to set up AI agent memory
Give your agent a folder of plain files it reads at the start of every session, then add a search index once that folder outgrows a single read.
Rules
How to write rules for AI agents
Put your agent rules in one plain file at the root of the project, usually AGENTS.md, keep each rule short and specific, and back the rules that matter most with a check that runs before the agent acts.
Content and media
24/7 AI video editor
How to set up an AI editor that cuts video around the clock: the steps, the checks it must pass, and the points where it waits for a person.
Tools and MCP
What is MCP
MCP, the Model Context Protocol, is an open standard that lets AI agents use your tools and data through small programs called servers. Which ones to connect first.
What are agent skills
Agent skills are folders holding a written procedure, a SKILL.md file plus any references and scripts, that an AI agent loads when a job matches. How to write one.
Models
Local vs cloud AI models
Send hard reasoning and customer-facing writing to cloud models; keep private data, offline work and bulk small jobs on a local model. A job-by-job guide.
Permissions
AI agent guardrails
AI agent guardrails are checks outside the model that stop a risky action before it runs. Which actions to guard first, and how to tell enforced from asked for.
Schedules and pipelines
How to schedule AI agents
To schedule AI agents, start each job from a scheduler that outlives the agent, make it safe to run twice, and have it tell you when it fails or goes quiet.
Watch and back up
How to back up and monitor an AI server
To back up and monitor an AI server: write, check, then swap each backup, keep one copy offsite, watch six signals, and stop big jobs before the disk fills.