FAQ

Short answers to the questions that come up most often: before installing Griglia, and in the first days of using the board. The long versions live in the pages each answer links to. If a word here is unfamiliar, the glossary defines it.

Before you install

Does Griglia work with my coding agent?

Griglia does not integrate with a specific agent: the contract is two Artisan commands and one instructions file. Any CLI agent that can run php artisan griglia:check and read a Markdown file can use the board — Claude Code, Codex CLI, Gemini CLI and others have been used this way. vendor:publish --tag=griglia-agents writes the instructions the agent reads. See the agent side.

Is it a separate application?

No. Griglia is a Composer package that installs into an existing Laravel application: composer require, php artisan migrate, and the board is served by your app, with your users and your database. See the installation tutorial.

Do I need an AI key or an internet connection?

No. The board, the agent workflow, notifications and themes work with no AI provider at all. Image descriptions, plans built from a prompt and voice transcription are optional and only run when laravel/ai is configured. See AI features.

Do I need Node, Vite or Tailwind?

Not in the default setup: Griglia ships precompiled CSS and JavaScript. If you prefer to compile the package with your own Tailwind build, set GRIGLIA_ASSETS=vite and follow front-end assets.

Is it stable enough to use?

It is used daily by its author, has a test suite running on every push, and follows semantic versioning — but it is still 0.x: minor versions may change behaviour, and the changelog says when. See the upgrade runbook and the governance and support policy.

Using the board

I added a task, but griglia:check says there is nothing to do

Two usual reasons, both by design:

• The task is still waiting. The agent only sees what you have released: tap the state dot once, so it becomes open to work. Everything else on the board is yours to edit while the agent stays out of it.
• The task is still waiting in another list. Only the agent list — agent_list in the configuration, dev by default, changeable with GRIGLIA_AGENT_LIST or in Settings — is the channel the agent reads top to bottom. In any other list of yours nothing happens until you release a task: mark it open to work and griglia:check shows it under 📋 List «…», after the agent list and after the plans. For everyday requests, keep using the agent list: that is where priority is set by drag and drop.

Started plans work the same way: their tasks show up after the agent list. See using the board and plans.

Can I run it without logging in?

GRIGLIA_MODE=local removes authentication and makes the lists global — for a board on your own machine, bound to 127.0.0.1. Never expose local mode to a network. On a server, access is controlled by canAccessGriglia() or GRIGLIA_ACCESS_GATE, and the administrative pages by canManageGriglia(), GRIGLIA_ADMIN_GATE or GRIGLIA_ADMINS. See access, administrators and modes.

Where does my data live?

In your application's database and storage: tasks, notes, attachments, questions, results, working time and token counts. Griglia sends nothing anywhere unless you configure notifications or an AI provider yourself.

Working with agents

Can more than one agent work at the same time?

Yes. Each list and each task can be assigned to an agent, griglia:check --agent=<key> shows only that agent's work, and a task claimed elsewhere is reported as busy. See two agents at once.

Can the agent keep working while I am away?

Yes, with a persistent worker: a small host process, published by vendor:publish --tag=griglia-scripts and usually run as a systemd user service, that polls the board and starts a fresh non-interactive agent session whenever a task is open to work. One instance per agent key, its own lock, and stop requests are honoured. See persistent workers.

How do I find out that the agent asked something or finished?

The board tells you by itself: in-app bell, Web Push on the devices you enable, and mail when a mailer is configured — with a deep link that opens the task. The two events follow the notify_on_done and notify_on_question settings, which silence both the board and the agent's own channel. See notifications.

What does it cost in tokens, and how do I spend less?

Griglia does not call any model on your behalf — the tokens are your agent's. It records what the agent reports when it closes a task and turns it into a cost on /stats, using the price per million tokens you set in Settings → App. To spend less, the ⚡ Optimization settings tune the agent's own habits: compact command output, terse mode, percentage updates only piggybacked on other commands. See statistics and the settings reference.