Pricing

Plans and Pricing

Start free, scale with usage.

  • Free

    $0 forever

    Try Schema on your own data.

    • One-time limited usage
    • Platform only
    • 1 seat
    • 50 MB storage
    • Docs and community support
    Start free

  • PLUS

    $50 /mo

    For individual builders.

    • Expanded usage
    • Platform and API
    • 1 seat
    • 5 GB storage
    • 2 concurrent batch jobs
    • Add-on usage
    • Email support
    Select plan
  • Most popular

    PRO

    $600 /mo

    For teams with production needs.

    • 15x usage
    • Platform and API
    • 3 seats
    • 50 GB storage
    • 5 concurrent batch jobs
    • Add-on usage
    • 99.9% SLA and priority email
    Select plan

  • Enterprise

    Custom

    For organizations with advanced needs.

    • Custom usage
    • Enterprise fine-tune program
    • Custom seats and storage
    • Custom rate limits
    • Encrypted (AES-256) model artifacts
    • 99.95% SLA and dedicated CSM
    Contact us

Pay only for what you use.

API

Operation Notes Rate
Schema Base Models Fresh cells, one rate for inputs and outputs $3.00 / Mcell
Cache Unchanged data on repeat calls $0.30 / Mcell
Batch Async, elected per job $1.50 / Mcell

Volume tiers apply to API inference. Contact us for volume pricing.

Operational

Synthetic data Per Mcell generated $1.50 / Mcell
Data storage Per GB / month, on storage overage $0.20 / GB

FAQ

  • How is usage counted?

    Usage is counted in cells. One cell is one row times one column, and a call’s cell count is the sum of each table’s rows times columns.

  • What does one pass return?

    Every pass returns the full bundle: sector identification, a per-column profile, predictions, missing-value imputation, and the cross-table map when the call spans two or more tables. Outputs are free on read operations; there is no per-output charge anywhere in the system.

  • Do missing values count as usage?

    Yes, at the normal rate. Schema encodes the pattern of missing values as signal and completes them; imputation is delivered work, not skipped compute. A null is a metered cell, not a discount.

  • How do weekly allowances work?

    Plus and Pro include a usage allowance that resets every 7 days and does not roll over. One allowance covers both surfaces: platform jobs and API calls made under a subscription key draw from the same bucket. Free is a single lifetime allotment with a hard stop. The pure pay-as-you-go API needs no plan at all.

  • What happens when my allowance runs out?

    Wait for the weekly reset, or add usage: prepaid credits metered at the public API rates, at par, with no markup. You can set a monthly spend cap, and alerts fire at 80% and 100% of your allowance. Free has no add usage.

  • How does cached pricing work?

    When you send the same data again, only what changed bills fresh: added and changed rows bill at the fresh rate, unchanged rows bill at the cached rate, and removed rows bill nothing.

  • What is batch?

    A service tier you elect per job at submit time. The job runs on a deferred queue and the whole job bills at half rate. A batch job that misses its completion window expires unbilled; you simply resubmit.

  • How are endpoints priced?

    An endpoint is your connected data plus configuration, created in minutes with a single encode; there is no training run. Requests bill as ordinary API inference at the same rates, and an idle endpoint costs nothing: the only recurring line is data storage.

  • Do failed jobs bill?

    No. A job either returns its full bundle or fails unbilled, and an idempotent retry draws only once. There is no per-request minimum anywhere.

  • Are there rate limits?

    Yes: per-organization request and cell-per-minute ceilings, plus batch concurrency per plan. They are anti-abuse speed caps, not usage buckets; exceeding one returns HTTP 429 with Retry-After. Enterprise contracts set custom ceilings.

  • Can I cancel anytime?

    Monthly subscriptions can be cancelled before the next billing cycle. Annual commitments include a 15% discount and run for the contract term.

  • How do enterprise contracts work?

    Custom usage, custom seats and storage, custom rate limits, the Enterprise fine-tune program, encrypted (AES-256) model artifacts, 99.95% SLA, and a dedicated CSM. Contact sales for a tailored proposal.