EigenLake

The agentic compute layer powering vector intelligence beyond search

EigenLake lets agents cluster, model topics, detect anomalies, and track change across large scale datasets.

Workloads

Run vector workloads where the data lives

Search, cluster, detect anomalies, and analyze trends over large-scale vector datasets without stitching together fragmented systems.

What Breaks at Scale

The hard part is the execution layer

Search returns neighbors. Pattern-finding across the full corpus still has to survive compute, permissions, metadata, retries, and a place to keep the answer.

Product question
Which warranty patterns changed this week?
idx.agent.query("find the warranty patterns changing this week")
Execution gap

No shared runtime owns the full query path.

Vector DB
neighbors only
Metadata
joins drift
Spark/GPU
batch job
Lambda glue
timeouts, retries
IAM
access gate
Result store
temporary output
The risk: analysis work moves from product logic into infrastructure glue.

Product

EigenLake: one Platform for all your vector needs

EigenLake gives developers one SDK to store vector records, search them, and run ML-style workloads on the same indexed data. EigenStore manages the vectors and metadata. EigenRun turns that stored data into search results, clusters, anomalies, topics, and time-series signals.

EigenLake SDK

connect()
index
records
workloads

EigenStore

The data service inside EigenLake. Create indexes, add records, attach metadata, and keep model-ready vector data in one place.

IndexesRecordsVectorsMetadata

EigenRun

The execution service inside EigenLake. Run nearest-neighbor search, clustering, agent queries, and higher-level workloads against stored records.

ClusterTime-SeriesAnomaly DetectionTopic Modeling
Store once. Search and run workloads without stitching together systems and workarounds.

Pricing

Same platform. Different scale.

Every plan includes the full EigenLake's platform: EigenStore for storage and metadata, and EigenRun for ML workload execution. The difference between plans is usage: stored vectors, search units, and executions.

Starter

$50 / month
Stored vectorsUp to 1M
Search unitsUp to 10k / month
ExecutionsUp to 10k / month
ComputeUp to 10 hours / month

Pro

$500 / month
Stored vectorsUp to 10M
Search unitsUp to 100k / month
ExecutionsUp to 100k / month
ComputeUp to 100 hours / month

Custom

Contact us
Stored vectorsCustom
Search unitsCustom
ExecutionsCustom
ComputeCustom

Try EigenLake free in the live sandbox, or schedule a call for production scale and deployment questions.

More details

Included in every plan

All plans include the same core product capabilities.

Vector dimensionsUp to 1024 dimensions
Metadata per vectorUp to 10 KB
Filterable metadata per vectorUp to 2 KB
Results per search unitUp to 100 results

Search unit policy

Each search unit returns up to 100 results. Requests returning more than 100 results consume additional search units in blocks of 100.

topK 1001 search unit
topK 2503 search units
topK 5005 search units
topK 1,00010 search units

FAQ

Questions about EigenLake

What is EigenLake?

EigenLake is a platform for storing vector data and running useful workloads on top of it. EigenStore gives embeddings, metadata, documents, and events a durable home. EigenRun lets teams search, cluster, model topics, detect anomalies, and analyze time-series signals against that stored data.

What workloads can run on EigenLake?

The landing page focuses on five workload families: search, anomaly detection, topic modeling, clustering, and time-series analysis. Search and clustering use the current client primitives directly, while other workload previews use agent queries until dedicated EigenRun APIs are available.

How is this different from a vector database?

A vector database is usually optimized for nearest-neighbor retrieval. EigenLake keeps that retrieval experience, then adds a unified data layer and execution layer for running higher-level vector workloads without copying embeddings into separate systems.

What is a search unit?

A search unit returns up to 100 results. If a request returns more than 100 results, it consumes additional search units in blocks of 100.

Can I try it now?

Yes. The sandbox lets teams try EigenLake in a live environment, and the docs cover the current Python client for connecting, creating indexes, adding records, searching, clustering, and using agent queries.

Schedule a call

See what your AI stack looks like when vector data and execution live in one system

Talk through your data, scale, and first workload with the EigenLake team.

Schedule a call