Skip to main content
These notes list the changes in each aiAxonIQ release that you can see or that change how the product behaves for you. Changes to our own operations — how we deploy, back up and monitor the service — are left out unless they affect you.

​
How releases work during the beta

aiAxonIQ is in private beta. Accounts are created by invitation rather than by public sign-up — see Create your account. Releases are named as release candidates for version 1.0.0: v1.0.0-rc.1, v1.0.0-rc.2 and so on. Each number is a later build than the one before it. The name says that version 1.0.0 itself has not been declared yet.
Your organization may not be on the newest release yet. The notes below describe what each release contains. They do not say when your deployment received it. If you need to know which release you are using, ask your account contact.
What is not listed. Some release candidates contained only internal changes and have no entry here. Work that has not been released yet is not listed until it is.

​
v1.0.0-rc.15 — 7 August 2026

Zero-code instrumentation, in the product

The eBPF zero-code agent — upstream OpenTelemetry eBPF Instrumentation, unmodified — is now part of the product:
  • A Zero-code page lists the hosts running the agent and shows how to install it. The list is built from the telemetry the agent already sends, so it includes agents installed before the page existed.
  • Get started offers zero-code as an install option, with the Docker and Kubernetes forms filled in with your endpoint and key.
Host health is worked out from how recently a host sent data, and the page says so. A host serving no traffic looks the same as one whose agent was removed. There are no start, stop or upgrade controls, because the agent has no control channel.The agent is early access. See Zero-code instrumentation.
The documentation was expanded to cover much more of the product. The API reference is now generated from the routes the API actually serves, with internal operations removed.

​
v1.0.0-rc.14 — 6 August 2026

One bad metric value no longer loses a whole batch

A metrics payload containing a single NaN or infinite value used to be rejected with HTTP 400, losing every other data point in it. Prometheus reports NaN routinely — for example, a summary with no recent observations — so pointing a Collector at Prometheus could fail outright.Now the non-finite values are dropped and the rest of the payload is stored. As little as possible is dropped: a bad quantile loses that quantile, not the data point. A missing value is not stored as zero, so a chart never shows a latency of 0 that was never measured.This applies to every path that accepts metrics, including Prometheus remote-write.

​
v1.0.0-rc.11 — 1 August 2026

Team invitations contain a working link

​
v1.0.0-rc.10 — 31 July 2026

Maintenance release. No change to product features.

​
v1.0.0-rc.9 — 31 July 2026

This documentation site

The first version of this documentation site: getting started, sending data with OpenTelemetry, the Collector, zero-code and Prometheus remote-write, platform guides, product guides and support pages.
Some public pages still named an old ingest address that no longer resolves. Every customer-facing page now gives the working endpoint. See Endpoints and errors for the address to use.

​
v1.0.0-rc.7 — 29 July 2026

Invitation and password-reset links

​
v1.0.0-rc.6 — 29 July 2026

Get started

An in-product setup flow: create a license key, find your OTLP endpoint, copy an install snippet for how you run software, and see your first telemetry arrive. Each step is marked done by checking the thing itself — a key exists, data has arrived — rather than by remembering a click. See Choose your onboarding path.
New organizations are created by invitation. The owner receives a link, sets a password, and lands in Get started. The link is handled in the browser, so an email security scanner that fetches it does not use it up. See Create your account.

​
v1.0.0-rc.5 — 29 July 2026

The playground

The public playground is complete: a workspace with telemetry already in it, to explore without an account. See Try it in the playground.

​
v1.0.0-rc.4 — 27 July 2026

Maintenance release. No change to product features.

​
v1.0.0-rc.3 — 27 July 2026

OTLP/protobuf metrics are accepted

Some valid OTLP metric payloads sent as protobuf — the default format for OpenTelemetry SDKs and the Collector — were rejected with HTTP 400. Because a 400 is not retried, those data points were lost. All five OTLP metric types are now accepted.

​
Next

Frequently asked questions

Short answers to the questions that come up most.

Troubleshooting

Find which stage of the pipeline a problem is in.