The Multi-Engine Data Warehousefor Agents & Builders

Query with SQL, run scripts, or give your agent secure data access via MCP.

codex mcp add oleander --url https://oleander.dev/mcp

Not just a warehouse. oleander brings Iceberg tables, serverless compute and intelligent routing together, so every workload runs on the right compute.

Iceberg Tables

Keep your data in Iceberg or connect the catalog you already use.

Right-Sized Compute

Serverless compute from single-node execution to distributed workloads, across multiple engines.

Intelligent Routing

Engine and capacity selected for every workload.

Built for agents

Secure access, governance, lineage and execution context.

Multi-engine. One platform.

Run SQL with DuckDB, DataFrames with Polars and distributed workloads with Spark. oleander handles the compute and traces every execution with OpenLineage.

View the Platform

Scripts

Run Python scripts on Spark or Polars.

my-spark-job.py

MCP

Give agents native access to your data with built-in routing and governance.

agent

SDK

Query and work with your data directly from code.

query.ts

CLI

Query and manage your data from your terminal or infrastructure.

CLI

Bring intelligent compute toApache Iceberg

oleander separates where your data lives from the compute that runs it, and picks the engine and capacity for each workload.

Try it with Snowflake Horizon

Bring compute to your data

Your tables stay where they are. oleander runs workloads against the same copy using the right compute.

Data Layer	Use
Managed Iceberg	Let oleander manage the data layer and secure agent access to your data.
Apache Iceberg	Connect the Iceberg tables and catalogs you already have.
Amazon S3 Tables	Run workloads directly against tables in your AWS account.
Snowflake Horizon	Query Iceberg tables through the catalog you already use.
External Databases	Sync MySQL, Postgres or MongoDB data into oleander in one click. No ETL required.

The right compute for every workload

oleander selects the engine and capacity at runtime.

Engine	Use
Bloom	oleander's native DataFusion engine for SQL, from single-node queries to distributed workloads.
DuckDB	Fast, lightweight SQL for interactive queries, mutations, and external data.
Polars	High-performance DataFrame execution for Python scripts, from single-node to distributed workloads.
Spark	Distributed compute for the largest data processing and Iceberg workload.

Built for agents

Teams build faster with AI-native tooling. oleander empowers your agents with infrastructure built for them.

Native Agent Plugins

Install our plugins for Claude + Codex so your agent can directly query your business data.

MCP

Connect our universal MCP to your favorite agent.

Guardrails

Prevent your agent from spontaneously clearing up a bunch of “free” space, cross joining all of your largest tables, and generally being a menace

Remediation Tooling

Tools like automatic anomaly investigation and instant RCA boost your agent.

Your agent sends a query

Use SQL or natural language to send a query to oleander.

Agent

MCP

/query

SELECT *

FROM events

WHERE country = 'US'

AND event_time >= NOW()

- INTERVAL '7 days'

oleander selects the engine and capacity

oleander's multi-engine router uses Iceberg table metadata, partitions, and dataset size to select the right engine and capacity, from single-node execution to distributed compute.

Picking engine...

Best fit

DuckDBCross-catalog & external data

SparkLarge distributed workloads

PolarsDataFrame workloads

BloomFlexible SQL execution

Target engine runs your query

Queries run instantly on the target engine on oleander's fully managed infrastructure.

Query run

Run ID: 7f3a2c...e91b04
Time: May 20, 10:42:31 AM
Engine: DuckDB
Duration: 2.34s
Cost: $2.10
Inputs: events
Outputs: events_us_7d

$2.10

event_id	country	event_time
e_1842	…	…
e_1841	…	…
e_1839	…	…
e_1833	…	…

Designed to Scale

Launch your oleander warehouse with a team of one, keep expanding it after you reach 1000 employees.

Launch your warehouse

Observe

Lineage. Robust lineage for all of your queries, datasets, and transformations powered by OpenLineage.

SQL telemetry. Query pre-structured telemetry data using SQL.

From the first query. Automated observability starting with your first query, so no history is lost.

Govern

Sensitive data. Lineage-powered governance makes tracing sensitive data easier than ever.

Environments. Multi-engine routing allows for easy separation of environments.

Permissions. Manage permissions on a user or agent level.

Grow

Engines. Expand your ecosystem with a virtually unlimited number of engines.

Compute. Configure heavier compute as your data grows.

External sources. Add a growing number of external sources, including Snowflake, to your oleander targets.

//Blog

Sep 28, 2026

How does this even work?

A visual walk through the query router's decision tree, and an agent simulator showing how an agent and oleander get to an answer together.

query-routingagentsmcp+5

PHPeter Hicks

Read

Sep 22, 2026

In AI We Trust. Everyone Else Gets Access Control.

AI agents need identities and least-privilege roles. How Oleander IAM enforces access control outside the model, from MCP tools to Iceberg namespaces.

iamaccess-controlagents+2

MPMinkyu Park

Read