LavaLake
lavalake.io
- Free plan
- Web
- API
- Self-hosted
What LavaLake is
LavaLake combines a data warehouse and lakehouse with SQL, dashboards, and AI agents for deployment in a customer’s own data center. Its assistant can locate data, query metrics, and create dashboards using the signed-in user’s permissions. An included MCP server lets external agents use platform tools under those permissions. Read-only database sources include PostgreSQL, MySQL, MS SQL Server, SAP Business One, and SAP S/4HANA; CSV and DATEV files can be imported, while PDF, Word, and text documents can be searched. Hybrid search combines vector and keyword results using a locally run nomic-embed-text model with PostgreSQL and pgvector. The lakehouse uses Apache Iceberg on S3-compatible storage, Nessie as its versioned catalog, and Trino as its query engine. Tableau and Power BI can connect through a read-only HTTP feed that applies analyst identity and records data fetches in an audit chain. The maker says lakehouse data and local search remain on the customer’s network, while external AI providers receive conversation content and assistant tool results. The free Core plan has no time limit and supports up to 16 vCPU on one machine. Paid plans start at €249 a month.
Who it suits
LavaLake suits teams that want SQL analytics, dashboards, and AI tools deployed in their own data center. Its access controls and data connections may suit organizations working with identity-aware analytics and varied database or document sources.
Good to know, and what to watch
- Assistant actions run under the signed-in user's permissions
- Imports CSV and DATEV files
- Searches PDF, Word, and text documents
- Connects Tableau and Power BI through a read-only feed
- Free Core plan has no time limit
- Streaming ingestion is not supported
- High availability is listed as planned
- Fully air-gapped operation is listed as planned
- Change data capture and continuous ingestion are in development
Plans
Every plan and priceCore
Team
Business
Enterprise
Free plan
Getting started
- Open https://www.lavalake.io/en
- Deploy the Docker Compose reference stack on one machine
- Use Ansible installation on Debian or Ubuntu if preferred
- Connect listed read-only databases or import CSV and DATEV files
- Configure S3-compatible storage for the Iceberg lakehouse
- Choose the Core, Team, Business, or Enterprise plan
Limits to check
Core costs nothing with no time limit, up to 16 vCPU, and a single machine. Team allows up to 10 users and 192 vCPU; higher tiers expand limits and support.
The verdict
LavaLake combines lakehouse storage, SQL, search, and permission-aware AI features, with a free single-machine tier and paid options. Note that streaming ingestion is unavailable and several deployment capabilities remain planned or in development.
Questions people ask
Does LavaLake offer a free plan?
Yes. Core costs nothing, has no time limit, and supports up to 16 vCPU on one machine.
Where can LavaLake be deployed?
It is designed for a customer’s own data center, with Docker Compose on one machine and Ansible installation for Debian or Ubuntu.
Which databases can LavaLake read?
Listed read-only sources are PostgreSQL, MySQL, MS SQL Server, SAP Business One, and SAP S/4HANA.
Can external AI agents use LavaLake?
Yes. Its MCP server exposes platform tools under the signed-in user’s permissions.
What does Team cost?
Team costs €249 a month, billed monthly, for up to 192 vCPU and 10 users.
What the maker publishes
- Access controls
- The security design uses Keycloak OIDC, RFC 8693 token exchange, Trino-enforced row access and column masking, and a WORM hash-chain audit log. source
- AI assistant
- Its built-in assistant can find data, query metrics, and create dashboards, with actions running under the signed-in user's permissions. source
- AI data handling
- The maker says the lakehouse and local search stay on the customer's network, while external AI providers receive conversation content and assistant tool results. source
- BI integrations
- A read-only HTTP feed supports Tableau and Power BI, applying the analyst's identity and recording data fetches in the audit chain. source
- Data sources
- The product lists PostgreSQL, MySQL, MS SQL Server, SAP Business One, and SAP S/4HANA as read-only database sources. source
- Deployment
- The reference deployment is a Docker Compose stack on one machine, with Ansible installation for Debian/Ubuntu; the Kubernetes Helm path has operational gaps including policy integration. source
- Files and documents
- It supports CSV and DATEV file imports and makes PDF, Word, and text documents searchable. source
- Founded
- 2026 source
- MCP
- The platform includes an MCP server that lets external agents use platform tools under the user's permissions. source
- Notable limitations
- High availability and fully air-gapped operation are listed as planned, and change data capture and continuous ingestion are in development. source
- Purpose
- LavaLake combines a data warehouse and lakehouse with AI agents, SQL, and dashboards for deployment in a customer's own data center. source
- Search
- Hybrid search combines vector and keyword results using a locally run nomic-embed-text model and PostgreSQL with pgvector. source
- Storage and formats
- The lakehouse uses Apache Iceberg on S3-compatible storage with Nessie as its versioned catalog and Trino as its query engine. source
Checked against the maker's pages on 7 Oct 2026.