New feature / Improvement

New values monitors, clearer answers for agents, and incident.io alerts

· week 47 · 2026

Weekly changelog card for W47 2026 listing new values monitors, clearer answers for your agents, monitor alerts to incident.io, starting to send data in one step, monitors that wait for a pattern, control over app access and SSO roles, and annotations on charts and deploys from CI
This week we shipped new values monitors, which tell you the first time something new shows up in your data, gave your agents clearer answers from the Axiom MCP server, and added an incident.io notifier for monitor alerts.

New values monitors

New values monitors watch one column of your query results and alert you the first time something new shows up: a new error message after a deploy, a new country someone signs in from, or a new tool your AI agent starts using.

  • Pick the column: Write an APL query, run it, and tick the column you want to watch.

  • Start with what's normal: With Learn from, the monitor records up to 30 days of past values as already seen. Forget values after sets how long they count as seen, 365 days by default.

  • Set a bar: Aggregate in your query to alert only once a value matters, such as a new client that sends at least 200 requests.

Each new value gets its own notification. Your agent can set these up too, through the MCP server or the Go SDK. For more information, see New values monitors.

Clearer answers for your agents

When your agent queries Axiom through the MCP server, every result now carries more of the context you'd see in Console.

  • Units and warnings: Columns are labeled with their units, and results lead with the same query warnings you see in Console, so an agent knows when rows were capped or estimated.

  • Empty or no match: The agent can tell a time range with no data apart from a filter that matched nothing. Paste a Console query link into the chat and it reruns that exact query, with its time range, from the qid in the URL.

  • Clean up fields, with your say-so: deleteDatasetFields removes unused fields from event datasets, but only after you tick a confirmation box.

Your agent can also set up email notifiers directly and hand you a Console link for channels that need a secret. updateDashboard is now replaceDashboard, and runSpotlight is now dimensionalAnalysis. Axiom skills also install in Amp now: clone the axiomhq/skills repo into your Amp plugins folder. For more information, see MCP server tools.

Send monitor alerts to incident.io

Monitors can now raise alerts in incident.io. Create an HTTP alert source in incident.io, then an incident.io notifier in Axiom with the alert source ID and token. Each group in a grouped monitor gets its own alert, and with Notify on resolve turned on, Axiom resolves the alert when the monitor recovers.

For more information, see incident.io notifier.

More of our favorite changes

  • Start sending data in one step: A new dataset page starts with OpenTelemetry, can create a scoped ingest-only token, and shows the ingest URL and your AXIOM_DOMAIN.

  • Monitors that wait for a pattern: Threshold and anomaly monitors now have Trigger after and Within the last. Match monitors accept lookup, parse-kv, union, and redact. See Match monitors.

  • Control app access and SSO roles: Connected apps on your profile lists MCP clients you've authorized; revoke one and access ends within five minutes. Admins set the default role for people who join through SAML SSO or Directory Sync.

  • Annotations on charts, and deploys from CI: Annotations now show on metrics and MPL dashboard charts, and you can choose which ones appear on monitor charts. Axiom CLI v0.22.0 can mark deploys from CI with axiom annotation.

  • Friendlier filters and query generation: In Query and Stream, whether you're querying events or metrics, the Add filter menu uses plain-language operators. Query generation picks MPL or APL from the dataset.

  • Around the Console: Span details has a Logs tab, events heatmaps draw one row per group (see Improvements to heatmaps), and Find trace works from any event with a trace_id.

  • More KQL in APL, and Go SDK v0.40.0: take_any, take_anyif, any, and anyif now work in APL. The Axiom Go SDK runs MPL queries and supports new values monitors.

  • See rejected data on the Usage page: Settings > Usage lists rejected data per dataset and reason. Ingest also rejects timestamps more than 24 hours in the future or before the Unix epoch. See Limits.