Ask Zero & AI insights

Zero is the Metryx AI. It answers questions about your community using the exact same live numbers your dashboard renders, so what it says always matches what you see on the charts. Chats are saved forever and shared with your profile's team, so anyone can pick a conversation back up later: the title in the top-left corner is the way back, and its dropdown lists every past chat by day. Each chat names itself after the first answer.

The picker under the question box chooses how Zero works on your next message:

• Auto (the default): Zero picks the right model for each request.
• Zero: the quick model. Fast, tight answers for everyday questions.
• Zero Analyst: deep analysis. It reads more of your data, compares windows, and takes its time.
• Zero Agent: actions. It asks what you want, then proposes a plan to build and configure your server.

Ask in plain language: "why did health drop this week?", "how many people were active in the last 30 days?", "who needs moderation attention?". Zero fetches the data it needs while it answers, and every fetch shows up under the reply as a source chip with the dataset, the window, and how long the fetch took.

Follow-ups work like a normal conversation. Asking "what about last month?" after a messages question re-runs the same lens on a different window.

Not sure where to start? A new chat opens with example questions you can fill with one click. And Zero can read your Metryx configuration too: ask about your AutoMod rules, leveling setup, access codes, scheduled reports, or who can moderate, and it answers from the actual settings.

When Zero reads your data, the answer comes with the picture. A question about activity renders a message chart right in the chat, a health question shows the score and its five pillars as a card, "when is my server active" draws the hour-by-day heatmap, and top inviters come back as a bar list. The charts are the same live numbers the dashboard shows, in a compact card under the source chips.

Zero can read everything the dashboard can: messages, member growth, voice, retention, text stats, moderation stats, invites, the activity heatmap, health, and the live snapshot. If a page can show it, you can ask about it and get the chart back in chat.

Zero does not just answer, it can do the work for you. Ask it to build out channels, set up roles, make a channel private, turn on welcome messages, add an AutoMod rule, schedule a report, and more. The rule of thumb: anything you can do on your dashboard or by hand in your server, you can ask Zero to do.

When you ask for something to be done, Zero replies with a plan: the exact actions it wants to take, in order. Nothing runs until you authorize it.

The moment Zero proposes a runnable plan, an Authorize plan popup appears above the question box: how many actions, on which server, with a "see steps" line that expands into the full list. Authorize once and the whole plan runs while a checklist works itself down the steps. Dismiss the popup and a small Run plan button waits in the plan's header instead.

Want changes first? Edit plan in the popup asks "how should this plan change?" — describe the tweak and Zero proposes a fresh plan that replaces the old one. You can also skip individual steps from the checklist before authorizing.

When a plan changes your server's structure, the plan card carries a View plan button. It opens a fullscreen preview of your server exactly as the plan would leave it: new channels marked, removed ones struck through, categories in place.

Use the Viewing as picker to browse the server as any role would see it. Channels that role cannot see disappear, and the message box at the bottom of each channel tells you plainly whether that role could post there — so you can confirm a staff channel is actually private before you authorize anything.

If the plan creates roles, a panel under the preview lists each one — open them one at a time to see the role's color and every permission it grants, in plain words.

Zero is built so it cannot run away, no matter what you or anyone else types. These limits are always on and cannot be turned off:

• Destructive work is gated. Any plan that deletes, bans, or locks makes you type your server's name before it runs, and large-scale deletions spell out exactly how much they remove.
• No admin powers. Zero can never create or hand out a role with dangerous permissions like Administrator, Manage Server, Manage Roles, Ban, or Kick. It will tell you to do that by hand in Discord.
• Content cannot trigger actions. If Zero reads member-posted content while helping you (names, messages, the audit log), every step of the resulting plan waits for your approval, so a message someone plants in your server can never make Zero act on its own.

Every plan Zero runs is recorded. Open Activity from the Ask Zero header to see a timeline of what Zero has done on your server, when, and under which trust mode, expandable to the individual steps and their outcomes.

Runs that can be reversed show an Undo button, both on the plan card right after it runs and later from the Activity page. Undo puts things back the way they were.

Each stats page opens with an AI Insights card: one sentence about the most notable change in that view, with the key number highlighted. The insight generates the first time anyone on the profile opens the page that day, then stays put until the next day in your profile's timezone.

The ring in the composer shows how much of the month's allowance the profile has used. Every model draws from the same allowance at its own rate — quick answers barely move it, deep analysis and Agent builds draw more. It resets on the first of each month.

The full story lives in Settings, AI usage: the meter, what each model costs to run, a plan-by-plan comparison, and what happens at the cap. Zero simply pauses when the allowance runs out; nothing is ever billed automatically.