An AI agent can now stand up a real Axiom organization on its own, work in it immediately, and hand it to a human to keep. Alongside that: multi-query charts for every query type, sortable legends and heatmap colors, zoomable traces, a new in filter for metrics, and a batch of improvements to the Axiom MCP server.
Agent-created organizations
An AI agent can now create a real Axiom organization on its own, with no signup and no human in the loop, then start working in it right away: creating datasets, ingesting and querying data, and building dashboards and monitors. It behaves like any other organization, not a sandbox.
The organization starts out temporary and lasts 24 hours. The agent gets back a claim link; open it, sign in, and the organization becomes permanently yours, with everything the agent built intact. Leave it unclaimed and it’s removed at the 24-hour mark. Until it’s claimed, the organization runs on a short-lived footprint (up to 10 GB of ingest and 1 GB-hour of query compute, notifiers paused, three organizations a day); claiming lifts those to the full Personal plan.
Point your favorite agent at the agent-created organizations documentation to get started.
Chart legends and heatmap colors
Borders, corner radii, dividers, and empty states around charts, query results, aggregation tables, and the Stream tab now come from one shared chart frame, so panels line up with each other wherever they appear. On top of that:
Aggregation columns in legends: Line and bar chart legends can show Last, Mean, Min, Max, Total, and Count columns for each series, so the numbers you’d otherwise hover for sit right next to the series.
Sort legends by any aggregation: Legends sort by the first aggregation by default. Click any aggregation in a right-side legend to sort ascending or descending by it; both right-side and bottom legends get a control to sort by any aggregation in either direction.
Bottom legends stay compact: A bottom legend holds its height to three rows and scrolls past that, so a high-cardinality series list keeps its hands off the plot area.
Higher contrast in dark mode: Axis labels and legend text in dark mode now carry the contrast light mode already had, which makes a dense time series easier to read across the room.
A color for heatmaps: Dashboard heatmaps gain the same Color setting statistic charts have, and cell values render as a continuous opacity gradient in that color rather than a fixed set of steps.
One refresh per dashboard: Auto-refresh runs every chart on a dashboard through a single batched request on one timer, so a refreshed dashboard is one consistent snapshot rather than a set of charts that each arrived at their own moment.
More ways to build charts
Multi-query and dual Y-axis for every query type: Multiple queries on one chart and a secondary Y-axis used to be metrics-only. They now work for structured, APL, and MPL queries alike, each query row with its own builder-or-editor mode and its own validation, so you can overlay related series and put two units on one chart whatever query wrote them.
Switch a chart’s type in place: A type picker in the chart editor moves a chart between time series, statistic, table, log stream, and the rest without recreating it, and keeps its limit, sort, and aggregation settings across the change.
Size metrics charts: Metrics queries get the same chart sizing controls as everything else on a dashboard.
Zoom into traces
Zoom and pan the waterfall: The trace timeline is now interactive: drag to select a range, use the mini-map, wheel-zoom, zoom to a span, or reset back out.
Honest gaps in the timeline: The event timeline in the Query tab dims the stretches where there are no events, so a gap reads as a gap rather than evenly spread data.
Markdown in annotations: Chart annotations render markdown, so a note on a spike can carry a link or a bit of structure. The Show annotations toggle also stays put on the query event timeline.
Query smarter with APL and MPL
An in filter for metrics: MPL queries take a
where … in [a, b, …]membership filter, with array values and array query parameters, so matching a tag against a set is one clause instead of a chain ofors.top after summarize in APL charts: APL dashboard charts allow
| topfollowingsummarize, so you can rank an aggregation directly in the chart’s query.Aggregation queries return more rows: A query ending in
summarizereturns a narrow result: its aggregations and group-by keys. Axiom now sizes the row limit to those columns instead of the limit it reserves for a query projecting every field, so many aggregations that used to stop at 50k rows now return more.Metrics query warnings surface: Warnings from an MPL query now show in the results banner, and a query whose aggregation changes a metric’s unit (like
map rate) shows a unitless Y-axis.
The Axiom MCP server
Dashboard sections: The dashboard tools understand sections, so an agent can create and edit dashboards that group charts into sections instead of one flat grid.
Correlation groups in listDatasets:
listDatasetsreports which datasets are correlated, so an agent knows which datasets it can join across before it writes a query.The APL skill loads inline: The
axiom-aplreference skill now loads into the session that asked for it, not a subagent, so its APL syntax is on hand the moment a query gets written. Install it withnpx skills add axiomhq/skills.
More of our favorite changes
The Where, Summarise, and sort-by field pickers now show a type icon and label (String, Number, or Timestamp) next to every option.
Each metric in an OpenTelemetry metrics dataset shows its unit, and the block visualization can list a dataset’s metrics as a table.
Changing an organization’s default retention lists the datasets it affects and asks you to confirm before a reduction takes effect.