Documentation Index
Fetch the complete documentation index at: https://docs.oleander.dev/llms.txt
Use this file to discover all available pages before exploring further.
What is oleander?
Every layer of your data infrastructure generates context. None of it connected. oleander changes that — unifying runs, datasets, schemas, queries, costs, traces, and logs into a single versioned context graph that engineers and AI agents can query directly. No new agents in your environment. No manual tagging. The graph updates on every run — always accurate, always current.Built by the co-creators of OpenLineage and Marquez. Backed by Julien Le Dem (Arrow, Parquet), Maxime Beauchemin (Airflow, Superset), Wes McKinney (Pandas, Arrow), and the developers behind SlateDB, and more.
Why oleander?
Your data stack already produces the context. Every tool sees a fragment. Nobody sees the whole picture. What hasn’t existed until now is a layer that takes all of it and turns it into one queryable context graph — for cost intelligence, optimization, and autonomous investigation.| Without oleander | With oleander |
|---|---|
| Warehouse costs spike. Nobody knows why. | Every query traced back to the pipeline and dataset that drove it |
| Three tools open to investigate one failure | oleander.query() across logs, traces, lineage, and cost in one call |
| Schema changes. Dashboard costs increase downstream. | Affected datasets and dollar costs surfaced before git merge |
| Agents act on incomplete context. Engineers get looped in. | oleander-mcp gives agents the same context your engineers use |
oleander collects telemetry via OpenLineage and OpenTelemetry — no direct access to your infrastructure required. Your metadata stays portable and on open standards.
How oleander works
Connect your stack
Route OpenLineage and OpenTelemetry at oleander from your existing tools — Airflow, dbt, Spark, Snowflake, BigQuery, and more. No agents, no direct access required.
oleander builds the context graph
Every run, dataset, schema, query, cost, trace, and log is unified, versioned, and connected into one living graph — updated automatically after every execution.