Pricing
Two contracts: Schift Context, then Schift Runtime
Schift Context makes your internal knowledge searchable, priced by scale. Schift Runtime turns your team's working sessions into a compounding asset on top of that Context, with governance and proposals included. Combine both, or start with either.
Schift Context
One monthly subscription for searchable company context
Search and Connect are included. Each plan also includes monthly ingest, on-demand compute without keeping Cold running, and AI Credit for generation and modification after search and Connect.
Starter
$100/moAnnual billing: $83/mo, $1000/yr
Search, Connect, ingest, on-demand compute, and generation/modification allowance are set from your active workspace entitlement.
Team
$350/moAnnual billing: $292/mo, $3500/yr
Search, Connect, ingest, on-demand compute, and generation/modification allowance are set from your active workspace entitlement.
Growth
$800/moAnnual billing: $667/mo, $8000/yr
Search, Connect, ingest, on-demand compute, and generation/modification allowance are set from your active workspace entitlement.
Business
$2500/moAnnual billing: $2083/mo, $25000/yr
Search, Connect, ingest, on-demand compute, and generation/modification allowance are set from your active workspace entitlement.
Contract 1 — the core
Schift Context — put your internal knowledge to work
Includes AI Search. Priced by tier, and every tier ships with a hard ceiling on search, ingest, and storage.
| Tier | Monthly | Max search | Max ingest | Storage |
|---|---|---|---|---|
| S | ₩500,000 | 60,000 쿼리/월 | 2,000 페이지/월 | 100GB |
| M | ₩1,600,000 | 220,000 쿼리/월 | 5,000 페이지/월 | 300GB |
| L | ₩5,000,000 | 660,000 쿼리/월 | 20,000 페이지/월 | 1TB |
| Enterprise | 문의 | 협의 | 협의 | 협의 |
Sizing assumption: 100 searches per person per day × 22 business days = about 2,200 queries per person per month. The contract carries the tier's limits, so 100 people searching 100 times a day and 200 people searching 50 times a day land on the same tier.
A tier is set by whichever axis you'd hit first — search volume or ingest volume, whichever is the tighter constraint for your team.
Going over doesn't stop anything
Cross a limit and you switch to metered overage automatically — we notify you at 80%, nothing cuts off.
| Axis | Unit | Price |
|---|---|---|
| Search overage | 10,000 쿼리 | $45 |
| Ingest overage — text | 1,000 페이지 | $28 |
| Ingest overage — scanned OCR | 1,000 페이지 | $92 |
We don't publish which model answers your searches — every tier runs on a lightweight model by design.
Contract 2 — on top of it
Schift Runtime — turn your sessions into a compounding asset
Governance and APM proposals ship inside this contract — collection, per-team limits, the spend ledger, and exposure policy all live here.
| Tier | Monthly | Compute VM | AI Credit | Weekly APM proposals | Weekly updates |
|---|---|---|---|---|---|
| S | ₩1,000,000 | 100시간 | 80cr | 8건 | 80건 |
| M | ₩3,000,000 | 300시간 | 250cr | 25건 | 250건 |
| Enterprise | 문의 | 협의 | 협의 | 협의 | 협의 |
AI Credit is the single consumption axis — proposals, updates, pack runs, and scheduled AI jobs all draw from the same pool. 1 credit = ₩2,000.
The weekly proposal and update counts are a velocity ceiling: they cap how fast the credit pool can be spent in a given week.
Go past the ceiling and you either buy Overage Credit or wait for next week's reset — which is free.
Governance is built into this contract — collection, per-team limits, the spend ledger, and exposure policy all live here.
Combinations
Most teams start with Schift Context and add Schift Runtime once they want their working sessions turned into a compounding asset.
| Combination | Monthly |
|---|---|
| Runtime S alone | ₩1,000,000 |
| Context S + Runtime S | ₩1,500,000 |
| Context M + Runtime M | ₩5,000,000 |
| Context L + Runtime M | ₩8,000,000 |
Schift Context comes first — it is the entry point, and Schift Runtime compounds on top of the Context it already built.
Prepaying annually gets you 10% off — one approval cycle a year instead of monthly billing.
Setup
Each contract has its own setup scope, priced separately after a conversation.
DX setup
Talk to usData ingest, migration, OCR, chunking, and permission mapping.
AI Asset setup
Talk to usBackfilling past sessions and running an initial optimization diagnosis.
Options — for extra protection
Only organizations that need these add them.
| Bring your own cloud (BYOC — AWS, GCP) | Available on any tier. Priced after we scope your headcount. | From $2,500/month · talk to us |
| Network isolation & regulatory compliance | Dedicated isolation — a separate vector DB and a full system migration. | Scoped by scale |
| Enterprise management | Extended retention · audit logs · SSO. | +₩300,000/month |
| Backup & disaster recovery | Redundancy · recovery-time commitments. | Scoped by scale |
How AI is billed — bring your own subscriptions and keys
You pick the AI. Connect the subscription or keys you already use, or run it through Schift's pay-as-you-go plan.
Connect what you already have — free
Bring your existing ChatGPT or Claude subscription, or a department API key, and connect it as-is. No extra charge — it never shows up as a line on our invoice.
Or use Schift's pay-as-you-go plan
We publish the operating fee: 10% for heavier models, 30% for lightweight ones. Every rate is public.
What building it yourself on AWS costs
For reference, here's what the same scale runs on raw AWS infrastructure plus the operating headcount to run it — before you've built any of the Korean-language handling or governance layer.
| Tier | Schift | AWS server cost | + operating headcount | Multiple |
|---|---|---|---|---|
| S | ₩500,000 | 178만원 | 338만원 | 3.6x |
| M | ₩1,600,000 | 303만원 | 463만원 | 1.9x |
| L | ₩5,000,000 | 527만원 | 687만원 | 1.05x |
Operating headcount assumes $800/month, and the initial build assumes 2 months at 1 engineer at $16,000 — separate from the monthly figures above.
Budget stops short on some of this: AWS Textract has no Korean support, HWP files raise an UnsupportedDocumentException, and connectors, ACLs, audit logs, and governance all have to be built by hand.
What we don't charge for
Your invoice carries three axes: search, ingest, and execution. The four below never appear on it.
Seats or headcount
Your employee count is a sizing input for choosing a tier — it never appears as a contract clause.
Accumulated memory
What's already stored never costs more, and we never ask you to delete it.
Memory call count
How many times memory gets read isn't metered separately from the search and ingest axes already in your tier.
The same resource, twice
Retrieval and APM runtime draw on the same compute — we never bill for it under two different line items.
What every contract carries
What we charge for is the infrastructure — search, ingest, compute, storage. The guarantees on top of it come standard.
Per-org data boundary
Org, bucket, collection, storage-prefix, and ACL boundaries are kept on every read, write, search, and execution path.
Usage visibility
Search, ingest, compute, and AI Credit usage are recorded line by line so you can see them.
Marketplace
A distribution fee on buying and selling APM packs. Early packs are supplied by us — we don't stand up a fee axis before there are third-party suppliers.
Frequently asked questions
Which contract should we start with?
Most organizations start with DX — it makes your internal knowledge searchable, and it's the entry point the Schift Runtime builds on. Add AI Asset once you want your team's working sessions turned into a compounding asset with governance included.
How is a DX tier chosen?
By whichever of the two axes — search volume or ingest volume — you'd hit first. We size search off roughly 100 searches per person per day across 22 business days. What the contract carries is the tier's limits.
What happens if we go over a DX tier's limits?
Nothing stops. You switch to metered overage automatically — we notify you at 80%. Search overage and ingest overage (text and scanned OCR are priced differently) each have a published rate.
What is AI Credit, and what happens if we run out?
AI Credit is the single consumption axis for the Schift Runtime — proposals, updates, pack runs, and scheduled AI jobs all draw from the same pool at ₩2,000 per credit. If you exceed the weekly velocity ceiling, you can buy Overage Credit or just wait for the free weekly reset.
Can AI cost more than expected?
AI is a pass-through. Connect an existing ChatGPT or Claude subscription or a department API key and there's no extra charge. Choose Schift's pay-as-you-go plan and the published rate is 10% for heavier models, 30% for lightweight ones.
Does prepaying annually lower the rate?
Annual prepay gets you a 10% discount on either contract. The underlying rates stay the same regardless of contract size.
Is setup billed separately?
Yes, and each contract has its own scope. DX setup covers data ingest, migration, OCR, chunking, and permission mapping. AI Asset setup covers backfilling past sessions and an initial optimization diagnosis. Both are quoted after a conversation — we don't publish a number without seeing your scale.
What if we need network isolation, regulatory compliance, or our own cloud account?
Bring-your-own-cloud (BYOC) starts at $2,500/month and is available on any tier. Dedicated network isolation is a separate option, scoped and priced by scale. Both are quoted after we understand your headcount and requirements.
Is this cheaper than building it ourselves on AWS?
At the scale our tiers cover, we come in well under a self-built stack — the S tier lands about 3.6x cheaper once you count operating headcount, and the gap narrows as volume grows. Some pieces sit outside budget entirely: AWS Textract has no Korean support, HWP throws an unsupported-document error, and connectors, ACLs, audit, and governance are all yours to build.
Want to scope this for your organization? Request a consultation →