01
Ask live data
Use plain English or SQL across connected databases, lakehouses, business apps, documents and streams.
Eligible paths avoid staging
The live intelligence layer
Ask every system.Trust every answer.
NeuronsDB is a federated AI query engine for governed, verifiable answers across live enterprise data. Query databases, apps, lakehouses and streams without centralizing them first.
Eligible zero-staging paths Predictive intelligence Proof Mode preview
Internal benchmark results · July 2026
Lower engine overhead. Matching results.
Measured against Trino across 247 matched workloads using the same fixture sources.
15.1×
faster p50
91.9%
lower peak memory*
247/247
matching results
246/247
p50 wins
Read the benchmark note
The 15.1× figure is the geometric-mean p50 improvement from the 28 July sequential local run against single-node Trino. NeuronsDB won 246 of 247 p50 comparisons with complete normalized-result parity. In the same run, maximum observed engine memory was 178.1 MB for the NeuronsDB worker versus 2,208.8 MB for the Trino container.
These are internal lab results, not a third-party certification. Memory is engine-only and source containers are excluded. The artifact does not embed hardware, cache-mode or environment provenance. Performance varies by query shape, source, scale and deployment.
Full methodology and raw resultsOne live intelligence layer
A query engine is only the beginning.
Traditional federation stops after returning rows. NeuronsDB can continue from a live question to an optional prediction, a decision and a preview proof artifact.
02
Predict what comes next
When configured, create forecasts, risk scores and recommendations connected to the data behind them.
Ready models work beside live data
03
Watch every change · Preview
Consume bounded stream batches, fold CDC events into queryable state and surface anomalies as data changes.
Streaming and CDC preview
04
Put answers to work
Deliver governed context to people, dashboards, agents and operational workflows from one place.
One connected answer layer
Proof Mode · Preview
Every answer can bring its receipts.
Successful planned responses can carry a signed record of the SQL, cited sources, lineage references and verifier status. An immediate result can also include an integrity check.
- Connect a successful plan to cited sources
- Detect changed or tampered proof artifacts
- Verify stored evidence for audit and review
NEURONSDB / PROOF 7F21 Preview proof
QuestionWhich enterprise accounts are likely to expand next quarter?
Answer
18 accounts show high expansion intent, led by product usage and support signals.
01
Sources5 live systems
02
VerifierSQL check passed
03
LineageReferences attached
04
IntegritySignature matched
Evidence chain intact
The integrated advantage
One runtime. Fewer systems to stitch together.
NeuronsDB combines the work usually split across federation, AI analytics, model operations, streaming and governance tools.
Capability
Federated query engines
AI analytics tools
NeuronsDB
Live cross-system analysis
Core
Connector-dependent
Built in
Forecasts and model lifecycle
Separate platform
Usually external
Configurable
Streaming anomaly detection
Separate platform
Usually external
Preview
Signed proof and lineage
External governance
Varies
Preview artifact
Private AI and air-gap path
No native AI layer
Often cloud-led
Self-hosted with air-gap preview
Category-level comparison based on commonly documented product scope. Individual products and deployments vary.
Enterprise control
Governance that travels with the question.
Query supported sources in place without making a central copy first. Policy, residency and evidence controls are available for supported paths, with some still in preview.
Policy-aware foundations
Tenant-aware policy, access, quota and sensitive-data primitives govern supported analytical paths.
Evidence in preview
Lineage metadata and signed proof artifacts can connect successful plans to cited sources and integrity checks.
Residency-aware admission
Define allowed regions and block disallowed analytical routes when source-region metadata is available.
Air-gap path · Preview
Self-host with local model backends today. Fully enforced, packaged air-gap controls are in development.
Deploy on your terms
From local services to your Kubernetes.
Run locally or self-host on Kubernetes in your private network. Local model backends are supported. A fully packaged air-gapped profile remains in preview.
Local services
Your VPC
On premises
Air-gap preview
Private preview
Bring your hardest data question.
We’ll run it across supported systems you already use and show the speed, resource use and available evidence side by side.