Pricing

Free to run. Help when you want it.

VeltrixDB is Apache 2.0 and self-hosted: you run it on your own hardware or cloud account and pay your provider for the machines. There is no managed cloud service today — one is planned.

Open source

For every team

The whole engine, on your infrastructure. No feature gates, no per-op meter.

$0Apache 2.0
  • Self-hosted: Docker, bare metal, Helm chart and Kubernetes operator (separate repos)
  • Standalone, replicated, or Raft cluster modes
  • Vector, BM25 full-text and hybrid search
  • AES-256-GCM at rest, RBAC, TLS / mTLS, audit log
  • 70+ Prometheus metric families, Grafana dashboards, alert rules
  • Community support on GitHub
Read the quickstart
Available today Engineering support

For production rollouts

Direct access to the engineers who build the engine, for your self-hosted cluster.

Contact us
  • Architecture and sizing review on your workload
  • Benchmark run on your hardware with our published harness
  • Migration planning from Redis or DynamoDB
  • Upgrade and incident help
  • Terms agreed per engagement
Book a demo ↗
You keep running open-source VeltrixDB on your own infrastructure
Managed service

Planned — not available yet

A hosted VeltrixDB we operate for you. It does not exist today; we will not take money for it until it does.

Planned
  • No launch date announced
  • No uptime SLA or compliance attestations today
  • Tell us your regions and requirements to shape it
Register interest
What you actually pay for

Your bill is your hardware.

With no licence fee, cost comes down to how much RAM and NVMe you provision. The difference from an all-in-RAM store is where the values live.

How to size a node before you price it

Plug your key count and value size into your cloud provider's price sheet. We will size it with you on a demo call — these are the rules we use, not a quote.

All-in-RAM storeDRAM
RAM ≥ data
Every key and value held in memory, plus per-key overhead. RAM grows with the dataset.
VeltrixDBNVMe
RAM ≈ index + cache
NVMe holds the values. RAM holds the index — ~142 B/key measured at 5M keys with the native index (since v1.6; 168 B/key with the Go map), plus ~90 B/key for the ordered index unless disabled — plus whatever cache you choose (default 256 MB). Using vector search? Add its index: ~3.4 KB per 768-dim vector as float32, ~1.1 KB int8, ~0.6 KB with PQ + disk graph (measured at 10K vectors).
Per-key figures are from an engine benchmark at 5M keys; measure with your own key lengths. No billion-key run has been published. Get a sizing review for your workload →

Rules of thumb are generic. Your workload isn't.

Get a TCO model on your real traffic ↗ What the bill includes
We size RAM and NVMe from your actual key counts and traffic · No commitment