RUMvision v2: marking releases via annotations, pipeline live within two days
RUMvision has just launched v2. And we at Elgentos, a customer since day one, were waiting for it eagerly. Not because of a new color scheme or a polished-up dashboard, but for one concrete reason: v2 comes with an API. And that opens the door to the kind of automation we've already been doing as standard for our clients for years with our other monitoring tools.
There's also an extra reason we're following this launch with a bit of extra pride: RUMvision, just like us, is from Groningen. Two Groningen-based companies working from the north on the performance of Dutch and international webshops. Groningen down-to-earth attitude, Groningen engineering, and no nonsense. It's this kind of cross-pollination between local players that shows you don't necessarily need to look to Amsterdam, Berlin, or San Francisco for world-class tooling. Stad en Ommeland, plain and simple.
Why annotations are the first thing we tackle
If you run a Magento or Hyvä store, there's one thing you don't want: not knowing where a performance degradation is coming from. Your graph dips on a Wednesday afternoon. Was that a release? A third-party script that quietly got 200ms slower? An A/B test that went live?
Annotations solve this by placing every deploy as a vertical line on your monitoring timeline. Do you see your LCP spike on Tuesday at 14:32, with a release marker sitting at exactly that moment? Then you immediately know where to look.
Our other monitoring stack, Tideways and Sentry, had already supported this way of working for a while. Elgentos's deployment pipeline automatically sends an annotation to both tools with every release, including commit hash, environment, and release name. RUMvision was the last link we still had to correlate manually. We're now putting an end to that.
Pipeline live within two days, rollout underway
We were prepared. As soon as v2 was released, we added the RUMvision endpoint to the same pipeline step that already calls Tideways and Sentry. One small piece of code, one secret in our CI, and a rollout plan we've already executed dozens of times.
The CI/CD integration was in place within two days. Every new client release that goes through the pipeline after that automatically sends an annotation to RUMvision. We're rolling it out to all existing client environments step by step, following each shop's regular release rhythm. No separate rush deploys, no ticket for the client, just piggybacking on the next release that's coming up anyway.
That's exactly why we place so much value on managing a monitoring stack centrally as an agency: one good integration, and all clients automatically benefit as soon as they go through the pipeline.
What you now see in practice
Correlation at a glance: if your Core Web Vitals score drops, you can immediately see whether there was a release right before it.
Faster rollbacks: if a degradation is confirmed against a specific release, the decision to roll back becomes trivial.
Better post-mortems: annotations give us a solid timeline to build on, instead of a reconstruction after the fact.
What we're building next
Annotations are step one. The next step is where it really gets interesting.
We want to lay all monitoring data over a single timeline — RUMvision RUM data, Tideways APM traces, Sentry errors, deploy events, business KPIs — and have that whole set analyzed by AI. Not to spruce up a nice-looking dashboard, but to answer the question that always costs time right now:
"Something's going on with this shop. What changed, and what's the most likely cause?"
An LLM with access to annotations, performance metrics, error rates, and conversion data can generate a hypothesis for that question in seconds, something that would normally take an engineer half an hour. That's the point where monitoring tooling tips from reactive to proactive.
A RUMvision MCP would be great
To take that next step faster, an MCP server from RUMvision would help enormously. MCP (Model Context Protocol) was introduced by Anthropic and is now supported by all major AI platforms. With a RUMvision MCP, any AI agent (Claude and our own agents) could directly and securely query RUM data, correlate it with other sources, and draw conclusions, without every party having to build its own bridge.
RUMvision has the data, we (and other clients) have the use cases. An MCP server would be the plug that connects those two worlds. Should this happen, we'd be happy to help think it through as a launching partner.
