MCP & AI
From question to dashboard
Query metrics and create dashboards in natural language. Your own agents can connect through MCP too.
- Choose your own model
- Changes require your approval
Cloud convenience. Your infrastructure.Data cloud for AI & analytics.
Data cloud for AI & analytics.
In your own data center.
Query your business data with AI, SQL and dashboards. Keep data, search and access permissions on your own infrastructure.
- AI
- Ask questions, find knowledge, create dashboards
- MCP
- Connect your own agents to your data
- 1
- Console for data, AI and governance
- 100%
- Your infrastructure for the data platform
AI in practice
What your assistant can do with lavalake
Ask questions. Put your data to work.
The assistant can explore data, query defined metrics and create dashboards. You work in natural language; tools execute each step under your identity. This example illustrates a possible workflow using a previously defined revenue metric.
Read the docs- 1
Ask a question
How is our revenue developing by region?
The assistant queries a previously defined metric.
- 2
Approve a dashboard
Create a dashboard from this.
Review the proposed change and confirm it.
- 3
Analyze together
One view for your team
Each tile runs under the permissions of the person viewing it.
The AI data platform
Built for agents, not retrofitted.
Connect data, find answers and control access in one platform.
Governance
Your permissions apply to your AI too
People and agents work under the same permissions. Access and changes remain traceable.
- Row filters and column masking
- Logged actions and data lineage
Getting data in
Make business data available to AI
Bring databases, files and documents into one lakehouse. Guided imports help you get started.
- Import databases, CSV, PDF and Word
- Continuous ingestionIn development
Knowledge & metrics
One knowledge base for people and agents
Use the same defined metrics across search, SQL, dashboards and AI answers.
- Local semantic search
- Shared metric definitions
Architecture
Cloud convenience for data and AI. On your infrastructure.
One web console for data imports, SQL, search, AI assistance and dashboards. Underneath sits an open Iceberg lakehouse with separate storage and compute. Run the platform in your data center, choose your AI model and keep your data in open formats.
The reference deployment runs as a Docker Compose stack on one machine; Ansible prepares installation and updates on Debian/Ubuntu. The Helm path for Kubernetes still has operational gaps, including policy integration. High availability and fully air-gapped operation remain planned.
Explore the architecture and technical details
01
Access
Console (SPA) · SQL worksheet · dashboards · AI assistant · MCP server — all behind one reverse proxy
02
Identity & governance
Keycloak OIDC · token exchange per RFC 8693 · row-access and masking policies · WORM audit · lineage
03
Compute
Trino under your identity · S, M and L workload classes in the Compose stack
04
Catalog & table format
Apache Iceberg with Nessie: ACID · time travel · branching · schema evolution · open to Spark, Flink, Trino
05
Storage & index
S3-compatible (MinIO, Ceph, NetApp StorageGRID, Dell ECS) · PostgreSQL with pgvector as the vector index
If you already know a cloud warehouse
The concepts stay, the bill changes. What is still missing says so.
- Proprietary table format
- Apache Iceberg with Nessie (open, versioned)
- Web interface for queries
- Console with SQL worksheet and data browser
- Roles and row filters
- Policies in Trino, under delegated identity
- Access history and provenance
- WORM audit log, lineage and access history
- Full-text and vector search
- Hybrid search with a local embedding model
- The vendor's AI features
- MCP server · the model of your choice
- Virtual warehouse
- Elastic compute clustersIn development
- Scheduled tasks
- Schedules for reports, threshold alerts and Iceberg maintenance
- Streams
- Continuous ingest and CDCIn development
- Credit billing
- A fixed subscription
Cost
More questions. More AI. No credits per data query.
Predictable platform costs instead of credits per query. See what you could save with lavalake on your own infrastructure, including hardware, power and operations.
€3,000€300,000
No credits
A fixed subscription based on vCPU and support instead of query credits. Queries from people and agents use your compute capacity without metering each operation.
No egress
lavalake charges no egress fees on your data. Budget separately for your infrastructure, network connectivity and connected providers.
Hardware you already own
lavalake runs on your existing Kubernetes and your S3-compatible storage. Capacity you have already written off keeps earning.
Planning you can budget
Cost follows cluster size and term, not usage behavior. Your teams are free to explore the data again.
Monthly comparison
81% cheaperCloud warehouse today€45,000
lavalake on-premises€8,768
- lavalake subscription
- €1,250
- Hardware amortization
- €3,500
- Operations & administration
- €1,913
- Power, including cooling and losses
- €812
- Storage, network, redundancy, backup
- €1,294
Model assumption: 6 servers, 48 vCPU · Business
Review assumptions and costs →Savings over 3 years
€1,304,343
Model of recurring costs. We calculate migration, parallel operation, floor space and AI models separately for your setup.
Request a TCO analysisSources for context
roughly half
Andreessen Horowitz estimates that repatriating cloud workloads cuts total cost of ownership to about half. An estimate rather than a survey, and calculated for software companies at scale.
a16z, The Cost of Cloud, a Trillion Dollar Paradox, 2021$2 million a year
37signals published the breakdown of its cloud exit: $3.2 million in annual spend, then roughly $2 million less per year after $700K of hardware. A single case with real invoices — and not a data warehouse.
37signals, Cloud Exit, 2024–2025
Sovereignty
AI on sensitive data. With clear boundaries.
Giving AI access to business data means controlling permissions and where data goes. In lavalake, people and agents share the same access rules. You choose which model to connect: self-hosted on your network or with an external provider.
You choose where your AI runs
- The lakehouse and local search stay on your network; external AI providers receive conversation content and tool results used by the assistant
- The vendor has no access to content or metadata
- The embedding model runs locally; semantic search needs no cloud key
- Fully air-gapped operation, including installation and updatesIn development
Provable, not asserted
- WORM audit log as a hash chain, over users and agents alike
- Lineage from source to reported figure, access history per table
- Every query under the identity of whoever is signed in — no shared service account
- Table and column classification with confirmed suggestions, without sending data values to the model
- Personal-data access, deletion and retention workflowsIn development
Operations
- Docker Compose for one machine, Ansible for installation and updates; Helm as the Kubernetes path
- Everything behind one reverse proxy; Trino and storage are unreachable from outside
- Daily database and configuration backup in the Compose stack; lakehouse files need separate backup
- High availability, a central observability stack and secrets through VaultIn development
Use cases
AI and analytics where your data lives
From business knowledge to answers
Assistants find documents, query curated metrics and create dashboards. Your own agents can access these tools through MCP — under the user’s permissions, with actions recorded in the audit log.
Search, metrics and dashboards through MCP
Machine data and ERP in one analysis
Production telemetry meets SAP orders — without mirroring sensor data into a public cloud. Scrap analysis runs on the same cluster as financial controlling.
Analyze production and ERP data together
Risk calculation with auditable provenance
Regulatory reporting and risk models read from versioned Iceberg tables. Time travel reproduces any reporting date, and lineage answers the supervisor's question.
Trace historical data snapshots
Clinical analysis inside your own network
Treatment and billing data stay in-house, with analysis and AI assistance running on top. Roles and masking make sure each department sees only its own slice.
Masking applies in the engine, not in the interface
One metric instead of four exports
Store data comes into the lakehouse from the upstream systems, and replenishment and marketing compute on the same curated metrics of the semantic layer instead of each on its own extract.
One metric definition for every report
Making case management systems analyzable
Data from separate case management systems is brought together without leaving the legal framework. Operated in the state data center, analyzed by self-service.
Operated in the state data center
Migration
Four weeks, and your cloud subscription is last winter's problem
Migration here does not mean rebuilding. Tables come over as Iceberg, roles and permissions move into Keycloak and the Trino policies. Both systems run in parallel until the results reconcile.
- Week 1
Taking stock
We read your existing warehouse metadata: tables, views, roles, query history and cost distribution. The result is a migration plan with a defensible TCO calculation.
- Week 2
Installation
lavalake is installed on your cluster via Helm chart and wired to your S3-compatible storage, your AD and your monitoring. No new hardware purchase required.
- Week 3
Data and pipelines
Tables are taken over as Iceberg and queries are re-pointed at Trino. Both systems run in parallel until the results match row for row.
- Week 4
Cutover
BI connections point at lavalake, roles and row filters are in place, reports are validated against the old figures. Then the cloud subscription gets canceled.
Pricing
A subscription, not a meter
The AI assistant, MCP server, search and governance are included in every tier. Pricing follows allocated vCPU and support, not the number of data queries. Budget separately for external AI providers or self-hosted models.
Your data lives on your storage. The subscription depends on allocated vCPU and support; data volume is not billed separately.
Team
€249
per month, up to 8 vCPU
For a department's first production warehouse: the complete feature set on a single machine, and someone who answers.
- Iceberg lakehouse on your storage
- SQL worksheet, import assistant and dashboards
- MCP server, hybrid search and metrics
- Up to 8 vCPU
- Up to 10 users
- Ticket support, answer within a business day
Growth
€549
per month, up to 24 vCPU
When the warehouse outgrows the department: three times the compute, and no user limit.
- The same feature set as Team
- Up to 24 vCPU
- Unlimited users
- Ticket support, answer within a business day
Business
€1,250
per month, up to 64 vCPU
For several departments on one platform: the cluster grows, the user count is open, the response time is committed.
- The same feature set as Team
- Up to 64 vCPU
- Unlimited users
- 8/5 support with a committed response time
Recommended
Enterprise
€2,499
per month, per cluster
For production use across several departments: unlimited compute, and tickets that are handled before everyone else's.
- The same feature set as Team
- Unlimited vCPU
- 8/5 support with a committed response time
- Prioritized ticket handling
Sovereign
Custom
Regulated environments
For organizations under special obligations: separated operation, offline updates, source code escrow and on-site support.
- Operation without an internet connectionIn development
- Secrets through Vault or External SecretsIn development
- Source code escrow
- On-site engagements
- Help with your certifications
60 million rows and around 50 simulated analysts on a machine with 6 vCPU and 24 GB RAM. Our documented load test shows what compact hardware can do. The measurements in detail
All prices net, excluding VAT. What counts are the vCPUs allocated to the platform's services, not your servers as a whole. There are no hosting costs — the platform runs on your infrastructure.
Further reading
Read the docs
DocumentationInstall it, understand it, run it
From the first start on your own machine through the concepts to the switches that matter in production.
15 pages
BlogArticles on cost, data sovereignty and MCP
Why cloud warehouses get expensive, what sovereignty means technically, and why agents need a semantic layer.
10 articles
CalculatorWhat your cloud warehouse would cost in your own data center
A model from disclosed assumptions — each with a source or explicitly labelled an estimate, all adjustable.
9 assumptions, sourced one by one
See AI work with your business data.
In 30 minutes, we show how the assistant finds data, queries metrics and creates dashboards. Together, we discuss deployment in your data center, your choice of model and the right access permissions.
- A live demo of the AI assistant and MCP tools
- Model choice, data flows and governance for your use case
- Next steps for a proof of concept in your environment