Pricing AI-ELT
Product - AI-ELT

An AI-native data layer, priced by what flows through it.

AI-ELT is Adya's data primitive - vector stores, embedding pipelines, agent memory connectors. Pricing scales with Monthly Active Records (MAR), the volume of records actually used in agentic workloads each month. Pilot at $1.5K. Hundreds of millions at $45K.

// AI-ELT tiers

Pick a tier by your monthly record volume.

A "Monthly Active Record" is any record that's been queried, embedded, or written by an agent in a given month. Idle records aren't counted - and aren't billed. Six tiers exist; below shows the four most common.

Pilot
1M MAR
$1,500 / month
Start with one workload at a million records monthly.
  • 1M Monthly Active Records
  • 50GB vector store
  • 10 embedding pipelines
  • Agent memory connectors
  • Standard embedding models
  • Email support
  • BYO embedding models
Start Building Run the calculator
10M MAR
Production
$10,000 / month
Most common tier for mid-market production workloads.
  • 10M Monthly Active Records
  • 500GB vector store
  • 100 embedding pipelines
  • Agent memory + custom connectors
  • BYO embedding models
  • Priority email + chat
  • Custom embedding namespaces
Start Building Run the calculator
30M MAR
Scale-out
$20,000 / month
Multi-region scale with custom embedding workflows.
  • 30M Monthly Active Records
  • 2TB vector store
  • Unlimited embedding pipelines
  • Custom + BYO connectors
  • Multi-region active-active
  • Dedicated CSM + Slack
  • Sovereign region pinning
Start Building Run the calculator
100M MAR
Enterprise
$45,000 / month
Hundred-million scale with custom retention and replication.
  • 100M+ Monthly Active Records
  • Unlimited vector store
  • Custom retention policies
  • Multi-region replication
  • BYO infra + on-prem available
  • Dedicated CSM, technical SE
  • Compliance: SOC2 / ISO 27001 / HIPAA
Talk to sales Run the calculator
Why MAR, not connector count?

Most ELT vendors price by source-connector count or row count. AI workloads don't fit either - a single agent might query the same 10K records hundreds of times. We bill unique records actually touched in a billing month. Untouched records sit free in cold storage; the meter only ticks when an agent actually uses them.

Read the design notes
// Capability matrix

Every data-layer feature, side-by-side.

In addition to the tiers shown, intermediate steps exist at 20M ($15K/mo) and 50M ($28K/mo) - the calculator routes you to the right one based on your declared volume.

Pilot (1M) 10M MAR 30M MAR 100M MAR
Monthly price $1,500 $10,000 $20,000 $45,000
Monthly Active Records 1,000,000 10,000,000 30,000,000 100,000,000+
Vector store size 50GB 500GB 2TB Unlimited
Embedding pipelines 10 100 Unlimited Unlimited
Agent memory connectors check check check check
BYO embedding models - check check check
Custom namespaces - check check check
Multi-region active-active - - check check
Sovereign region pinning - - check check
Custom retention policies - - check check
On-prem deployment - - - check
Compliance attestations - - SOC2 SOC2 / ISO / HIPAA
Support Email Priority + chat CSM + Slack CSM + technical SE
// 30M MAR - annualised

AI-ELT 30M vs splitting it across vendors.

Adya AI-ELT 30M
$240,000 / year
30M MAR - 2TB vector - custom retention - multi-region - SOC2 ready
Pinecone Pro
~$120-360K / year
Vector store only - ~$10-30K/mo at 30M scale - no embedding pipelines, no ELT
Fivetran
~$216K / year
~$18K/mo - ELT only - no vectors - no agent memory - still need a vector store
Self-managed (Postgres + pgvector)
~$80-200K / year
Engineering cost - 1-2 FTE on infra - ongoing scaling and replication
AI-ELT positions as an AI-native combination of ELT and vector primitives, not a like-for-like replacement of either category. See the full comparison matrix.

Stand up the data layer your agents actually need.

AI-ELT scoping calls take 30-45 minutes. We map your existing data sources, estimate your MAR, and propose a tier with 20-30% headroom for growth.

Start Building Run the calculator