Compose investigation pipelines.
Where, summarize, project, extend, join, take, sort, count, distinct, parse, lookup. The operators most ad-hoc log investigations need. Tab-completion in the Axiom Console; copy-paste examples in /docs/apl.
PLATFORM / QUERY
APL: piped, sequential queries at petabyte scale.
Axiom Processing Language is piped and log-friendly, built for the workflows engineers run every day. Power users get logs, events, and traces in APL, metrics in MPL. Schema-on-read, MCP-queryable.
CUSTOMERS RUNNING APL AT SCALE
Power users, picking it up in an afternoon.
Dashboards in days, not weeks
Built custom dashboards on production data in days. Pipeline composition felt like SPL they'd graduated to.
A year of Lambda logs, queryable
Queries a year of AWS Lambda logs — 4 billion messages — in APL, with no sampling and no rehydration step.
Vercel-edge visibility in days
Stood up edge-function visibility across the Vercel fleet in days, querying every request in APL.
ARCHITECTURE / HOW APL WORKS
Piped, sequential, schema-on-read. Designed for the questions you actually ask.
APL is piped. You compose a query as a sequence of operators (where, summarize, project, extend, join) that flow data left to right, the same way engineers think when they read a log file. Engineers from SPL and SQL backgrounds pick it up in an afternoon.
TECHNICAL DEEP DIVES
Everything you reach for, in one language.
Transform fields without re-ingesting.
Virtual fields let you compute, normalize, or extract values at query time. Define once, use across queries, dashboards, and monitors. Ingest-time schema decisions don't have to be perfect.
Try APL on live data, no account needed.
Open the Playground at /play. Useful for cheat-sheet conversions from SPL or SQL, and for testing patterns before they hit production dashboards.
Schedule queries that drive monitors and tiles.
A scheduled APL query becomes the source of a monitor, a dashboard tile, or a downstream system. The same query language for ad-hoc work and production alerting.
Query metrics alongside logs and traces.
MPL handles metrics aggregation; APL handles logs, events, and traces. The Axiom Console renders both side by side, and dashboards mix MPL and APL panels in one view.
Let your AI agents query the same data.
The native MCP server exposes APL to agents with cell budgets and schema introspection sized for token budgets. No glue code or per-vendor adapter.
“Coming from KQL, Axiom's query language felt very natural.”
— Hydroxygen Labs
FAQ
What engineers ask first.
LANGUAGE COMPARISONS
APL is the closest-in-class language to SPL today: piped, sequential, log-friendly. Most query patterns translate one to one. A cheat sheet at /docs/apl/spl-to-apl covers the operator mapping.
APL is piped instead of nested. For sequential, log-style analysis (filter, then summarize, then project), APL reads top to bottom where SQL nests subqueries. Joins, aggregates, and window operations are all supported.
APL is heavily inspired by KQL and shares much of the syntax. If you're coming from KQL, the transition is usually a day.
DATA MODEL AND SCALE
Metrics aggregation is a different kind of workload than log-style filtering, with pre-aggregated time-series and high-cardinality tags. MPL is purpose-built for it. Keeping the boundary explicit keeps both languages clean.
Yes, in the same APL query. Logs, events, and traces share EventDB and APL has join operators that work across datasets.
Petabyte-scale. Hapn keeps a year of AWS Lambda logs queryable (4 billion messages); other customers run multi-TB-per-day workloads.
AGENTS AND LEARNING
Yes. The native MCP server exposes APL to agents using the same primitives engineers use, so an SRE agent can run the same patterns a human runs in the Axiom Console.
Start with /docs/apl. The Playground at /play runs APL against live data with no account. The Axiom Console has tab-completion and inline help.
One query language. Every signal. APL on top.
Petabyte-scale ingest. Pricing you can predict.