Every month, someone on your team asks the same question: "how's everything actually doing?" Answering it means clicking through the endpoints list, the traffic dashboard, the firewall and rate-limit pages, the usage meter, and the certificate expiry table — then pasting it all into a doc nobody wants to write. An ai account report collapses that into one command. You ask Iris, the AI that runs inside the Ngris gateway, to produce an account report, and it reads your real account data with your own forwarded token and hands back a structured, table-per-area document you can export. It reads. It never writes, and it never fabricates a number to fill a gap.
The monthly "how's everything doing" question
Open an authenticated chat — the streaming widget or the full-page chat — and ask in plain English:
# in an authenticated Iris chat
Give me an account report for the last 30 days.Iris runs the account-report playbook: a read-only flow that chains together the same account-scoped reads you'd do by hand — endpoints, traffic, usage and billing, policies, certificates — and assembles them into one document. Because the reads run under your forwarded token, the report is scoped to exactly what you can already see and nothing more. Guest chat can't do this; it has no account to read. There's no setup, no new API key, and no separate reporting service to stand up.
What's in it
The output is structured markdown, one table per area, built entirely from real reads:
- Endpoints — your live and configured endpoints, their hostnames, and which region serves them.
- Traffic — request volume, status-code distribution, and top paths over the window, pulled from your real traffic records.
- Security posture — firewall rules, rate-limit rules, and traffic policies currently in effect, plus how they're covering recent requests.
- Usage — bandwidth and request counts against your plan limits, and current billing state.
- Certificates — issued certs, their hostnames, and expiry dates so a wall-clock surprise doesn't take a service down.
Each section is a table you can drop straight into a status doc or a ticket. If a section has no data — no rate-limit rules yet, no certs issued — Iris says so. It does not invent a plausible-looking row to make the table feel complete.
Exportable and read-only
The document Iris hands you is exportable to CSV, HTML, or PDF — the CSV for a spreadsheet, the HTML or PDF for something you can attach to a monthly review. The export is a rendering of the exact data that was read; there is no separate "generate an impressive export" step that could drift from reality. And the whole flow is read-only by construction. Producing a report is a diagnosis operation, not a mutation, so it never touches your configuration.
That matters because on Ngris, AI writes are gated in a way reports simply don't trigger. When Iris does change something — say you asked it to draft a firewall rule — every write surfaces a confirmation card with a field-level before-and-after diff, and firewall, rate-limit, and traffic-policy writes additionally get a blast-radius dry-run that replays the change against your real recent traffic before you approve. A report performs none of that. It's reads all the way down, and every AI tool call — read or write — lands in the AI-activity ledger, so "what has the AI done on my account?" always has an honest answer.
How it works
The report is a playbook: a fixed, read-only sequence that chains individual account reads into a single document. Iris doesn't improvise which admin endpoint to hit — playbooks call the same account-scoped read APIs the dashboard uses, under your token, and admin and internal APIs are hard-denied to Iris regardless of what any prompt asks. So the report can't reach into another tenant, can't escalate, and can't rewrite anything on the way through.
If a section reveals something worth acting on, the report is a natural jumping-off point into the rest of Iris. A spike in blocked traffic pairs well with per-request inspection and replay; a gap in coverage is a cue to have Iris draft a firewall or rate-limit rule — which then goes through its own confirmation-and-dry-run path, not silently. The report tells you where to look; the generators and the confirmation gate handle the actual change.
If you're already running endpoints on Ngris, the account report is one message away. Get set up with brew install ngris, point a tunnel at a local service:
# start an endpoint, then ask Iris for a report
ngris http 3000 --url app.ngris.comThen open the chat and ask for your account report. The free plan gives you 5 endpoints with no credit card, which is plenty to see the whole flow end to end — read your account, build the tables, export the doc, and never once fabricate a row.