Pay only for what you use.
No plans, no tiers, no usage caps. Top up your prepaid USD balance, and pay only for the operations you use.
How charges are calculated
Every billed operation uses the same formula.
Provider cost is the price Anthropic and OpenAI publish per token type — input, output, cache write, and cache read. The pricing multiplier varies by operation type — see the exact multiplier for each one below.
What gets billed
Four operation types are billed, each with its own pricing multiplier. Everything else — uploads, dashboard usage, and API calls that don't invoke a model — is free.
Agent chat
Claude Sonnet 4.6
Every chat message the agent answers, from the first input token to the last output token.
Document vision (OCR)
Claude Haiku 4.5
Runs only for scanned documents or files without a text layer. Vision-based extraction processes each page as an image.
Document indexing
OpenAI text-embedding-3-large
Runs for every ingested document, with or without OCR, to make it searchable through semantic search.
Semantic search
OpenAI text-embedding-3-large
Runs whenever a chat message triggers a search — a small charge before the agent generates a response.
Provider costs by model
Provider costs per million tokens, as published by Anthropic and OpenAI.
| Model | Input | Output | Cache write | Cache read |
|---|---|---|---|---|
| claude-sonnet-4-6 | $3.00 | $15.00 | $3.75 | $0.30 |
| claude-haiku-4-5 | $1.00 | $5.00 | $1.25 | $0.10 |
| text-embedding-3-large | $0.13 | — | — | — |
The pricing multiplier is applied automatically based on the operation type — see the exact multiplier for each one above. The balance deducted from your account already reflects the final amount charged.
Illustrative example
Rounded example values to show how pricing works. Your actual cost depends on document size and conversation length.
Uploading a 10-page scanned PDF
Document vision
Claude Haiku 4.5 processes the scanned pages as images (~8,000 input tokens) and extracts the text (~2,000 output tokens).
$0.05403× multiplierDocument indexing
The extracted text (~2,000 tokens) is converted into embeddings for semantic search.
$0.00083× multiplier
One chat message with search
Agent chat
Claude Sonnet 4.6 reads the retrieved context and conversation (~3,000 input tokens) and generates a response (~500 output tokens).
$0.04953× multiplierSemantic search
The question (~30 tokens) is converted into an embedding to search your knowledge base.
< $0.00013× multiplier
Frequently asked questions
Is there a plan or a usage cap?
No. There are no tiers, seats, or monthly limits. Pricing is entirely based on actual usage, with no artificial caps.
What happens when my balance runs out?
New operations are blocked until you top up your balance. The operation that exhausts your balance still completes successfully — only subsequent operations are blocked.
How does the pricing multiplier work?
The pricing multiplier is applied to the published provider cost and varies by operation type. You can see the exact multiplier for each billed operation in What gets billed above.
Do I need a credit card to start?
No. Every new account starts with $1 in free credits. No credit card required.
Turn your documents into answers today
Create your free account and turn your documents into answers in minutes.