Status
Last updated 2026-08-14 · Artifinance is operated by Anthrotech · contact hello@artifinance.io
All systems operational — as of the date above
This page is a plain statement of what's up, what we're targeting, and what's actually watching the service today — not a marketing claim dressed up as a dashboard. Where something is manual rather than automated, we say so, because a due-diligence reviewer or a customer during an incident finds the gap either way.
Current status
All systems operational, confirmed manually as of the "last updated" date above. This status is reviewed and republished by hand today — it is not yet a live feed. Treat the date at the top of this page as the freshness of the information, the same way you would with any status page that hasn't moved to real-time yet.
Separately, the in-app apps hub shows a status indicator inside the product. It was previously a fixed label; it now reads a real signal. It says "All systems live" only when tenant-book durability is configured and the background sync loop has actually started, and otherwise says "Running — books not backed up". It fails closed — if the check itself cannot run, the indicator keeps the honest not-backed-up wording rather than turning green. It is deliberately narrower than this page: it reports the durability posture of the instance serving you, not reachability, latency or model availability. This page remains where we make the service-wide statement.
Agent workforce self-test
Artifinance's pitch is an agent workforce whose wiring is machine-checked rather than asserted, so the check itself is published here. Every agent on the roster declares the tools it can call; the self-test resolves each of those names against the application's live dispatch map and downgrades any agent whose tool no longer resolves. It runs on every load of the in-product workforce page, and the figures below are regenerated from the same registry code — they are not typed by hand, and a drift between them and the registry fails our build rather than sitting here quietly.
Roster self-test: passing — 146 of 146 tool-holding agents verified, 0 unverified
- Roster — 159 agents across 13 departments.
- App wiring — 524 of 524 declared tools resolve to a real, callable function in the application's live dispatch map.
- Unverified agents — 0. An agent is unverified when it claims a tool that no longer resolves; it is shown as “defined, not yet surfaced” in the product rather than as a live agent.
- Last verified — 2026-11-13, the date these published figures were last regenerated from the registry.
What this does not claim: that every agent has been exercised end-to-end on your data, or that a resolving tool is a correct one. It proves the wiring is real, which is the part vendors in this category most often assert without evidence. If the self-test cannot run at all in a deployment, the product shows "could not run" rather than a green result, and we treat that as a platform incident.
Availability target (SLO)
We target 99.5% monthly uptime for the hosted app at app.artifinance.io, measured against a successful response from the health endpoint. This is a target we're holding ourselves to going forward, not a historical average — we don't yet have an automated uptime history to report a real measured number against, and we'd rather state that plainly than publish a percentage we can't back with data. As the monitoring described below moves from configured-but-unapplied to live, this section will report real, measured numbers instead of a target.
Incident severities we use internally:
- Major outage — the app is unreachable or core workflows (sign-in, ledger writes) are failing for most or all customers.
- Degraded performance — the app is reachable but slow, or a non-core feature is failing.
- Maintenance — a planned deploy or migration with a known, bounded window.
What's monitored today
- An external uptime check against
https://app.artifinance.io/api/health, checked every 60 seconds from outside Google Cloud — the one check that can tell "the service is fully down" apart from "no visitors right now." - Alert policies for elevated 5xx error rate, p99 latency, internal-error log volume, and durability-sync failures.
- A Cloud Monitoring dashboard covering request volume, latency percentiles, error rate, and instance count.
All of the above are real, reviewable configuration in this repository's monitoring/ directory. Honestly, as of this page's last update, none of it is applied yet — it is not yet applied to the live Google Cloud project, and no alert is yet wired to a paging channel (email/SMS/Slack) — so today, a real outage is first noticed by a human checking, or by a customer reporting it, not by an automated page. Applying that configuration and wiring a paging channel are the two concrete steps that would make this page's "as of" caveats above go away.
Planned maintenance
Nothing is scheduled. There is no planned maintenance window open at the time of this page's last update.
Routine releases do not need one: a deploy goes out as a new Cloud Run revision and traffic is switched to it once it is serving, so the previous revision keeps answering until the new one is up. Work that genuinely needs a window — a data migration that has to pause writes, for example — is announced in this section before it starts, with the date, the expected duration, and what will be unavailable, and is updated here again when it is finished. If you are running a pilot, this is the section to check before a period close or a filing deadline.
Incident history
No incidents have been recorded on this page since it was published (2026-08-13). This is a fresh page, not a claim of a long clean track record — check back as it accumulates real history.
Report an issue
If something looks down or slow, email hello@artifinance.io with what you're seeing and when — we'll investigate and update this page.
Who maintains this page
Artifinance is operated by Anthrotech, and this page is owned by whoever is operating the service — reachable at hello@artifinance.io. That matters because a status page with no owner is the one that goes stale first. Concretely: the availability, monitoring, maintenance and incident sections are written and reviewed by hand, and carry the date at the top of this page as their freshness; the agent-workforce figures are generated from the code they describe and are guarded by an automated test, so they cannot silently drift from the product even between hand reviews.