Tutorial 03

Take Auxot's pulse in 10 seconds

Open System Health and learn to read the dashboard at a glance — your daily ten-second confirmation that everything's running smoothly.

Plus: three prompts that turn this whole dashboard into a one-paragraph briefing your Admin Agent gives you instead.

Audience	Everyone
Time	~3 min
Prerequisites	An Auxot account (any tier). Helpful but not required: you've finished Tutorial 01, so the Admin Agent feels familiar.
You'll end up with	A 10-second daily check you can do whenever you want a pulse — plus three Admin Agent prompts that turn the dashboard into a quick oral briefing.

Why this matters

System Health is the page that confirms everything’s working — your AI models, your agents, your integrations, anything connected to your Auxot account.

The whole point is one place to look. When something feels off, you don’t have to guess whether it’s you, the AI model, the agent’s setup, or your internet — you check the dashboard and you have your answer in ten seconds.

Most days, you’ll glance, see green across the board, and move on. (Auxot keeps working whether you check or don’t.) On the days something needs attention, this is where you find out fast — usually before anyone else asks why the AI is being weird.

Today, learn to read it. Tomorrow, when you want a pulse, you’ll know exactly where to look.

System Health gathers the status of everything connected to your Auxot account — AI models, agents, integrations, recent activity — into one view that updates live.

Quick start

Sign in — open Auxot in your browser and log in.
Open System Health — click System Health in the left menu.
Read the top card first — that’s Platform Status. A green dot and “All systems operational” is your one-glance all-clear.
Skim the cards below — each one tracks one kind of thing connected to your account (your AI models, your agents, your integrations, etc.). Anything that’s “Online” or “Healthy” is fine. Anything else is worth a closer look.
Glance at Recent Activity — the running log on the right shows what’s been happening lately.

Done? You’ve done your first 10-second pulse check. The rest is habit.

The agent can do that?

You don’t always need to read this page yourself. The Admin Agent can read it for you and brief you in a paragraph. Three prompts that turn the dashboard into a thirty-second oral briefing.

1. Find what’s quietly drifting

Look at System Health and the recent activity stream. Find what's quietly drifting — a queue that's been deep for hours (Auxot won't fail, but I might want more capacity), an integration that keeps flapping on and off, a provider getting close to its quota cap, anything still online but heading the wrong way. Tell me what to watch and what to do, in priority order.

Why it’s non-obvious: Auxot is built so LLM work doesn’t fail — it queues. That’s a feature, not a bug. The trade-off: the system absorbs spikes silently, which means the early signals of “you’re outgrowing your setup” or “an integration’s about to drop” get absorbed instead of screaming red on the dashboard. The Admin Agent reads the patterns that mean you should make a quiet decision now — add capacity, refresh a credential, top up a quota — before users start feeling the wait times. Catching it now means a calm decision instead of an angry one.

2. Compare today against last week

Compare today's System Health and recent activity against what they looked like one week ago. What's worse? What's better? What's different in a way I should care about? Be specific and brief.

Why it’s non-obvious: A snapshot of right now tells you the current state. The Admin Agent can hold two timelines in its head — last week’s and today’s — and tell you what’s actually moved. Things you’d otherwise have to keep mental notes on (or scroll through three browser sessions of activity logs to compare) become a single paragraph: here’s what’s better, here’s what’s worse, here’s what’s different that I should care about.

3. Draft the customer-facing status update

Look at System Health right now. If something's degraded or down, draft me a short, honest update I can post to our customer Slack channel (or status page, or email blast) — what we know, what we don't, what we're doing, when we'll update next. If everything's fine, tell me so and stop there.

Why it’s non-obvious: When something’s wrong, your customers want to hear from you fast — and you want to sound calm, specific, and competent. That’s a hard combination to nail in five minutes. The Admin Agent reads what’s actually happening in the dashboard, writes the kind of update your most thoughtful colleague would write if they had thirty minutes free, and hands it back to you to skim, edit, and post. Stop drafting from scratch under pressure. Start editing instead.

Go deeper

What “Online” actually means on each card

The word “Online” means slightly different things depending on the card, and the difference can matter:

Model Providers — “Online” means the AI service is reachable and accepting requests. The provider is ready to handle whatever your agents ask of it.
Connected Agent Filesystems — “Online” means the agent’s separate workspace is reachable. The page itself notes: “Online / offline is the auxot-agent filesystem connection only — not chat availability.” In other words, an agent can be “offline” here and still totally fine to chat with.
Tool Workers — “Online” means the worker process is connected and ready to handle tool calls.
Integrations — “Connected” means the link between Auxot and the external service (Slack, Discord, etc.) is established.

Empty cards aren’t broken — they’re just unused. Each card has helper text inside showing the exact next step when you’re ready to set one up.

Troubleshooting (when something’s not green)

Two ways to handle anything that’s not green: try the concrete first step below, or screenshot the offending row, hand it to the Admin Agent in chat, and let it walk you through. The agent is usually faster when the error message doesn’t mean anything to you yet.

Platform Status is “Degraded” or “Down” — the database (Postgres) or cache (Redis) that Auxot runs on isn’t reachable from the Auxot server. This isn’t fixable from inside the Auxot UI — it’s a server-side issue. If you run the Auxot server yourself, screenshot the Platform Status card and ask the Admin Agent: “Platform Status is showing degraded — walk me through what to check on my server, step by step.” It can tell you which service is the likely culprit (database vs. cache vs. network), which command to run to confirm, and what the output should look like when things are healthy. If someone else runs the server, screenshot the card and send it to them — they’ll know which service to inspect.
A Model Provider says “Offline” or “Error” — usually authentication or a network issue. Click the row to open that provider’s settings; the error message there is more specific than what you see on the dashboard. If the message doesn’t immediately make sense, screenshot both the System Health row and the settings page error, then ask the Admin Agent: “This provider is showing offline — here’s what I’m seeing. What’s the fix?” It can usually point at the cause and the next step.
A Model Provider says “Over quota” — you’ve hit the usage limit on that provider’s plan (your OpenAI account is rate-limited, your Anthropic plan is out of credits, etc.). Three options: wait it out (rate limits reset on a clock), switch to another connected model in the meantime, or top up that provider’s account directly. Want help triaging? Ask the Admin Agent: “My [provider name] is over quota. Look at the work that’s currently queued and waiting on it — what’s urgent and what can sit? Based on that triage, tell me whether to wait it out, switch the urgent items to another connected model, or top up that provider’s account. Help me decide what to do next.”
An agent or worker shows “Offline” — usually the worker process isn’t running, or its connection has dropped. If you set the worker up yourself, restart it (Tutorial 10 covers worker setup if you need a refresher). If someone else manages your workers, screenshot the offline row and pass it along — or run it past the Admin Agent first: “This [agent or worker] is showing offline. What’s the most likely cause, and is there anything I should check before flagging it to my admin?”
Recent Activity is empty — not a problem. The feed is tied to your current browser session, so a fresh tab always starts empty. Do anything in Auxot (send a chat message, change a setting, open another page) and the feed will populate.

Variations & edge cases

Brand-new account: Most cards will be empty on day one. That’s normal. The page becomes more useful as you connect more things — Tutorials 04 (context files), 06 (teammates), and 09 (AI models) all add to what shows up here.
What you see depends on your access level. On Free tier (solo), you see everything on your account. On Team, Business, or Enterprise, you see what your role lets you see — your own activity, your team’s connections, your team’s recent jobs. If a teammate is operating at a scope you don’t have visibility into, their activity won’t show up here. The same scoping applies when the Admin Agent reads System Health on your behalf — it inherits your access level, so the briefing it gives you reflects your view, not the whole company’s.
A few reads need org-admin permission specifically. The job queue, the detailed provider status, and other people’s chat threads are all gated to org admins. If you’re a regular User and you ask the Admin Agent for any of those, it’ll tell you it doesn’t have access — that’s expected, not a bug. The remaining commands (your own agents, your context files, recent events, your own threads) work fine for any user.
Free vs. paid tiers: The page shows the same six cards regardless of tier. Platform Status shows your license info — Free tier shows a “no commercial license” note; paid tiers show your tier name and organization.
Live updates: The page updates in real time. You don’t need to refresh — if a worker connects, a provider goes offline, or an agent does anything, the relevant card updates within a second or two.
Queue-depth bars are relative. The little bars next to each Model Provider compare against the highest load any provider has reached during this browsing session, so the bars rescale as new peaks happen. They’re a visual sense of “who’s busy compared to whom right now,” not absolute job counts.
Mobile view: On phones, the cards stack vertically and Recent Activity moves to the bottom.

Walkthrough

Open Auxot in your browser and sign in.

Step 2: Open System Health

Click System Health in the left menu. The page loads with a row of cards on the left and an activity stream on the right.

The page’s own helper text describes it well: “Live status for core services, model providers, agents, tool workers, and integrations. The activity column streams events from your current browser session.” That’s the whole page in one sentence.

Step 3: Read the top card first (Platform Status)

The card at the top is Platform Status. It rolls up the health of Auxot’s core into one indicator:

Green dot, “All systems operational” — everything’s fine. This is what you’ll see almost every day.
Yellow dot, “Degraded” — one underlying service is acting up. Auxot still mostly works, but something needs attention soon.
Red dot, “Down” — both core services are unreachable. This is rare, and the fix is on the server side.

This card also shows your license tier — Free, Team, Business, or Enterprise — so you can confirm at a glance which plan this account is on.

Tip: If Platform Status is yellow or red and you don’t admin the account, the fix is usually on whoever does. You’ve still done your job by noticing — flag it and move on.

Step 4: Skim the cards below

Below Platform Status, you’ll see four more cards that each track one thing connected to your account:

Model Providers — the AI models you’ve connected (OpenAI, Anthropic, your own GPU, etc.). “Online” or “Healthy” means they’re ready to handle requests. The little bar next to each provider shows how busy it is right now.
Connected Agent Filesystems — only relevant if some of your agents have their own dedicated workspace set up (most agents don’t, and that’s fine). “Online” means that workspace is reachable.
Tool Workers — extra capabilities you’ve plugged into your agents — things like web search, code execution, or custom tools your team built. “Online” means they’re ready when an agent reaches for them.
Integrations — apps like Slack or Discord connected to your agents. “Connected” means messages flow between Auxot and that channel.

Anything that says “Offline,” “Error,” or “Over quota” instead of “Online” is your signal to look closer. Click any row to jump straight to that thing’s settings, where you can see why it’s offline and (usually) fix it.

Tip: When something’s not green and you’re not sure how to fix it, take a screenshot of the row, open chat with the Admin Agent, and ask it: “What does this status mean and what’s the next step?” It can usually walk you through it from there.

If a card is empty, that just means you haven’t set up that kind of thing yet — for example, no Slack integration means an empty Integrations card. Empty isn’t a problem; it’s just “not used.”

Step 5: Glance at Recent Activity

The right column is a live log of what’s been happening — sign-ins, agent jobs, providers connecting and disconnecting, anything else worth recording. Newest events at the top.

You’re not meant to read every event — you’re meant to glance. If you see lots of red ✗ marks or the same thing disconnecting repeatedly, that’s the kind of pattern worth investigating.

What’s next

→ Tutorial 04: Add your first context file — once you know everything’s healthy, give your agents the business knowledge that makes them seriously useful.
→ Tutorial 05: Give your agent its job description — if an agent’s behavior feels off (rather than its connection), the fix is in its job description, not in System Health.
→ Tutorial 09: Connect a cloud AI model — if your Model Providers card is empty, this is the tutorial that fills it.

Reference

Pages in Auxot: System Health, Settings → Providers, Settings → Agents, Settings → Tool Policies
Updates: Live (no refresh needed)
Card-by-card: Platform Status, Model Providers, Connected Agent Filesystems, Tool Workers, Integrations, Recent Activity
See also: Tutorial 09: Connect a cloud AI model