The AI Toolkit is the Weblabor coding standards, the agents and commands that apply them, and the analyzer that checks them — served to your AI client over MCP.
You never clone it. You connect your assistant to the server once, ask it to install the toolkit, and from then on it reads the rules when a task needs them and sends the files it writes to be checked.
The problem it solves
A standard somebody is expected to have read is a standard that goes unfollowed. Your assistant writes code all day and has no way of knowing how your team writes it, so what comes back looks like everybody's code except yours.
The toolkit turns the rules into a check that runs. You send a file, and you get back what to review, the line it is on, and the reason the rule exists. People and AI agents are held to the same rules: there is no stricter version for the machine and no relaxed one for the human.
What you get
| What it gives you | |
|---|---|
| The standards | Coding rules for Laravel, Livewire, Blade, Tailwind and Alpine, published in English and Spanish. Your assistant reads them on demand. |
| The agents and commands | Specialists and workflows — planning, review, refactor, documentation — installed into your own AI client in the shape that client expects. |
| The analyzer | A check that reads one file and reports what it can prove, with the line the problem is on. |
| Your projects | The plans, analyses, product vision, pending findings and internal commands of each repository, kept on the server and read in the Projects section. See Your projects and their documents. |
The standards and the guides are never copied into your project. They are read from the server every time they are used, so a rule changed today is the rule every project follows tomorrow, and you have nothing to pull. The agents and commands are copies, so those you reinstall when they change — see Install the toolkit.
What it does not do
- It does not run formatters. Mechanical formatting stays with your project's own tooling, and the assistant is told not to fight it.
- It does not read your disk. Every check carries the path and the content in the request; the server never opens a file on your machine.
- It does not give you anonymous access. Every call carries a token, and every run is recorded under the account that made it.
- It does not replace your own reading of the code. An assistant saying a file looks fine is not an analyzer pass, and the toolkit reports the two as different things.
Where to start
- Get your access token — create an account and generate the token every call is signed with.
- Connect your client — Claude Code, Codex, Cursor and four more, each with its own guide under Your AI assistant.
- Install the toolkit — the agents and commands land in your AI client, once.
Then you are working: analyse a file, read what comes back, read the standards, and see what each command does.