AI Work OS: run your AI agents like a small, careful company

What it fixes

AI agents can research, write and build all day. The hard part is everything around the work: knowing what the agent actually did, stopping it before a command you cannot undo, and picking up tomorrow where you left off today.

AI Work OS is the layer I run my own work on, cleaned up so you can copy it. It is a folder of plain files plus a few small scripts. No server, no account, no subscription.

How one job moves: you ask for an outcome, the rules and the board steer it, the agent builds, proves it and writes a receipt, and you approve anything that sends, spends or deploys

What a normal day looks like

Morning: "what now". You open Claude Code in your workspace and type two words. Before the agent answers, a start-up script prints the top of your board: today's focus, what is waiting on you or on others, and your lanes (a lane is one project or area of work). The agent picks one outcome instead of asking you to explain everything again.

A new session printing the board's "Context for next session" block and the list of lanes

Midday: hand a job to another agent. Say "write a job packet for GPT to add a contact form; it may only change the contact page folder." You get a short file in dispatch/inbox/ with the goal, the exact files it may touch, the checks that mean "done" and when to stop and ask. Then open GPT's Codex CLI in the same folder and say "run the packet in dispatch/inbox". GPT works from that file, not from a long chat. When it finishes, it leaves a receipt: what changed, what was checked, what is still unproven.

Any time: the agent tries something you cannot undo. Say the agent decides to force-push, which overwrites the shared copy of your code on GitHub and can wipe other people's changes. A safety check reads the command first, blocks it and tells the agent to ask you. The same check lets harmless commands through, even a commit message that mentions a dangerous command.

The safety check blocking a force push and telling the agent it needs the human's approval

Evening: "eod". The agent writes a short log of the day ending with the exact next action, and updates the board if anything changed. Tomorrow's "what now" starts from there.

Three words you will see

Board. One file, memory/TODO.md, with one short row per outcome: who owns it, where it stands, the next action.
Receipt. The note an agent leaves after a job. If there is no receipt, the job is not done.
STATUS line. Every "done" reply ends with VERIFIED, UNVERIFIED or BROKEN, plus the checks behind it. It turns "trust me" into "here is what I ran".

Start in 20 minutes

You need Claude Code, git and python3 on a Mac or Linux computer. On Windows, use WSL. The kit is free; Claude Code or GPT's Codex CLI may need a paid plan. v2.3 has 40 skills in ten packs:

aiwos-core: onboard, what now, eod, job packet, handoff, evidence loop, and a business brief that every job reads.
aiwos-business: sort leads, draft WhatsApp replies, build quotes from your price list, check a message before it goes out. Drafts only; a person always sends.
aiwos-sales: find prospects from public information, plan a discovery call, write a scope-led proposal (prices only from your own rate rules, with the maths shown), and keep a follow-up queue.
aiwos-delivery: client kickoff, produce the deliverable, and a separate quality gate that says pass, revise or blocked.
aiwos-plan: a weekly bottleneck review and a decision stress test.
aiwos-handoff: hand work between AIs (to GPT, to Claude Code, or to a chat AI like ChatGPT), receive it, check what came back against the real files, and get a second AI to review it.
aiwos-marketing, aiwos-web, aiwos-video: Google Business posts and follow-ups, a build-and-check loop for web pages, and video briefs with a beta caption checker.
aiwos-thinking (optional): seven thinking tools.

Most new jobs come in two editions: one for Claude Code or GPT's Codex CLI, and a copy-paste prompt (PROMPT.md) you can use in any AI chat. The repo also has a one-person weekly rhythm (workflows/weekly-rhythm.md) that chains the jobs Monday to Friday, and a guide to using more than one AI in the same workspace (docs/multi-agent.md).

Create a new workspace (start here). This makes your rules file, board and safety hook, and adds nine packs (33 skills) to this project only:

TEXT

git clone https://github.com/Rebelzxr/ai-work-os.git
cd ai-work-os
./setup.sh ~/my-work --link-skills
cd ~/my-work
claude

Say "onboard", then "what now". Run ./doctor.sh ~/my-work from the kit folder any time to check what is linked and wired. After the first setup finishes, run ./setup.sh ~/my-work --link-skills --pack thinking from the kit folder to add the seven thinking tools. Using GPT's Codex CLI? Add --codex; the automatic safety hook runs in Claude Code only for now.

Customer data: start the customer-drafting skills with made-up examples. Before you paste anything real, remove names, IC numbers, phone numbers, bank details and addresses yourself, in your own editor. The safe-to-paste skill is a second check after that, not a replacement for it.

Add skills to an existing workspace (plugins). Plugins add skills only; they do not create the rules file, board or hooks, so use the route above if you don't have those yet:

TEXT

claude plugin marketplace add Rebelzxr/ai-work-os
claude plugin install aiwos-core@ai-work-os --scope project
claude plugin install aiwos-business@ai-work-os --scope project

Plugin skills are called with the pack name in front, for example /aiwos-core:what-now.

Already on v1 or an earlier v2? Update your clone with git pull, then run ./setup.sh ~/my-work --upgrade first. It only shows what would change, file by file, and writes nothing. Review the differences and merge wanted changes into your existing files. Then run ./setup.sh ~/my-work --link-skills to add missing files and skill links while keeping your existing files. Links from older global installs stay until you remove them. Only use --force if you really want fresh copies: it backs up every file it replaces into the workspace's .aiwos-backups/ folder and prints each path.

What it will not do

The safety check is a careful pattern check, not a sandbox. It catches the common forms of dangerous commands, not every trick. The repo lists what it misses.
The scripts run in Claude Code. GPT and Cursor follow the rules file, but nothing stops them automatically.
A STATUS line records what was checked; it cannot prove the check was right. For anything public or about money, have a fresh agent review the actual files.

Where I use it

The library you are reading, the daily AI news and every release of this site go through the same board, packets, receipts and STATUS lines. Independent reviews kept finding ways to slip commands past the safety check; each one is now among its 213 tests.

The per-repo layer underneath it is Set up a project for Claude Code: memory, permissions, checks and safe autonomy. The owner-only actions it protects are the same five as The five things AI should never do without you, written for a business rather than a repo. Repeated hand-offs and checks become one-command skills with Write a Claude Code skill for a job you repeat: page captures with error checks. The same read-only review idea, applied to one decision instead of a release, is A decision board: six read-only AI reviewers for one hard call, and the news job that reports into this board is described in A daily AI news digest researched by GPT, with links it cannot invent. For one agent inside your own program, with its tool limits written in code, see Claude Agent SDK: when to use it, with a tested example.