Pricing

Pay for bytes stored, bytes written and queries: no servers, no seats, nothing for idle time.

Free
$0
Storage1 GB
Written1 GB / month
Queries100,000 / month
Document writes100,000 / month
Namespaces1,000
Query CPU1 hour / month
Cold object-storage reads1 million / month
Data returned10 GB / month
Hosted embeddings5 million tokens / month

No card. Past the limits, requests are refused, never billed.

Pay as you go
$10.00minimum / month, counted against usage
Storage$0.10 / GB-month
Written$0.25 / GB
Queries$0.50 / million
Cold object-storage reads$0.50 / million
Hosted embeddings$0.02 / million tokens

Unlimited namespaces and keys; no caps on CPU or data returned. Upgrade from your dashboard.

Cost calculator
Storage$0.10 / GB-month
1M docs (~4.6 GB)
$0.46
Writes$0.25 / GB
10% rewritten a month (~407 MB)
$0.10
Queries$0.50 / million
1M queries a month
$0.50
Estimated cost $10.00 monthly minimum, counted against usage
$10.00/month

Cold queries also pass on their object-storage reads ($0.50 per million reads; a warm query makes none). Idle namespaces cost only their bytes.

Compare the same workload on turbopuffer and Pinecone →

Two worked bills

At the prices above (October 2026). Stored bytes are what SkimmerDB keeps: compressed text is about half of what you send.

A SaaS with 200 customers, one namespace each
Storage2M documents of 1 KB: 1 GB$0.10
Written0.4 GB a month$0.10
Queries3M a month$1.50
Cold reads200k a month$0.10
Hosted embeddings50M tokens$1.00
Usage$2.80
Billlarger of usage and the minimum$10.00
Agent memory for 20,000 users, own embedding key
Storage50M memories: 25 GB$2.50
Written10 GB a month$2.50
Queries60M a month$30.00
Cold reads5M a month$2.50
Embeddingsyour provider bills you$0.00
Usage$37.50
Billlarger of usage and the minimum$37.50

Questions

What counts as a query?

A search, vector search, count, document read, warm-up or SQL query. A SQL query is one query however many namespaces it joins. Listing namespaces and managing keys are free.

What do embeddings cost?

With your own OpenAI, Voyage, Gemini or Cohere key, nothing from us: the provider bills you. Hosted embedding (voyage-3.5-lite) is $0.02 per million tokens, the provider’s price with no margin; the free plan includes 5 million tokens a month.

What is a cold read?

When a namespace isn’t cached on a server, a query reads what it needs from object storage. Those reads are passed through at cost. Warm queries make none.

What does an idle namespace cost?

Only its storage. Nothing runs for a namespace nobody is using.

What happens past the free limits?

Requests are refused (HTTP 402, saying which limit) until the next month or until you upgrade. You are never billed on the free plan.

Why does the free plan count CPU and bytes returned?

So the free plan stays free for everyone: 1 CPU-hour of queries, 1,000,000 cold reads and 10 GB of answers a month cover far more than 100,000 ordinary queries need. Pay as you go has no such limits.

Why is my free allowance smaller before I verify my email?

Until the account owner verifies their email, a free account gets 1% of each allowance, so signups work at once but throwaway accounts can’t use the whole free plan. Verifying lifts it.

How does the minimum work?

Pay as you go has a $10.00 monthly minimum. Usage counts against it: you pay the larger of the minimum and your usage, not both.