Billing
Hindsight Cloud offers two billing models: Pay-As-You-Go (self-service) and Enterprise Billing (invoiced). View your balance, purchase credits, manage payment methods, and review invoices from the Billing page.
Accessing Billing
- Log in to Hindsight Cloud
- Open the Organization menu in the top navigation bar and select Billing
Owners and Admins have full billing access (purchase credits, manage payment methods, configure auto-recharge). Members can only view credit balance and usage.
Credits Balance
The Credits Balance card shows your current account status:
- Current credit balance (in dollars)
- Effective rate for each operation (per million tokens, per call, or per million tokens per month for storage)
- Estimated capacity remaining by operation type
Pricing
Usage is measured per operation. Some operations are charged per million tokens (the volume you send or receive); others are charged a flat fee per call; storage is charged per million tokens stored per month and billed daily.
| Operation | Unit | Rate1 | Description |
|---|---|---|---|
| Retain | per million input tokens | $10.00 | Storing memories |
| ↳ Import | per million input tokens | $5.00 | A subcategory of Retain: importing documents from an export archive, or restoring a bank from one. No fact extraction runs, so it costs less than retain. Listed under Retain in the Billing page's Pricing card, and counted as Retain usage |
| Recall | per million output tokens | $0.75 | Retrieving memories |
| Reflect | per call | $0.05 | AI-powered insights |
| Retrieve Model | per million output tokens | $0.25 | Fetching a mental model, or reading a knowledge base page (lightweight lookup) |
| Refresh Model | per call | $0.05 | Creating or refreshing a mental model, or creating or updating a knowledge base page |
| File Conversion (Iris) | per million output tokens | $7.50 | Enhanced extraction of uploaded documents (PDF, images, Office docs) |
| Storage | per million tokens / month | $0.25 | Memories retained longer than 30 days |
A knowledge base page is stored as a mental model, so knowledge base activity is billed on the two mental-model lines above rather than under a separate heading:
- reading a page bills as Retrieve Model
- creating or updating a page bills as Refresh Model, at $0.05 per call
Pages also refresh themselves by default. A knowledge base page is created with
trigger.refresh_after_consolidation enabled, so after the consolidation that
follows your retains, any page whose scope actually received new memories
re-synthesises — and each of those is another Refresh Model call. Pages that saw
nothing new are skipped. (Mental models you create directly are the other way
round: that trigger is off unless you turn it on.)
So this line can grow with how much you write rather than with anything you did
deliberately, and in a bank with several broadly-scoped pages under steady
retains it can end up larger than every other operation combined. If it looks
bigger than you expected, the per-operation breakdown on the Billing page shows
the call count behind it. Setting trigger.refresh_after_consolidation to
false on a page makes that page refresh only when you ask it to. See
Retain for the mechanism.
Give a model or page a schedule rather than refreshing it on every
consolidation: trigger.refresh_cron caps it at the ticks you choose, and a tick
whose scope saw nothing new is skipped rather than billed. Refreshing on
consolidation has no ceiling — it follows your write volume, so the same page
costs more in a busy week for reasons unrelated to the page.
Automatic refreshes are also rate-limited to at most one rebuild per model per minute, so a burst of retains bills one call rather than one per retain. Refreshes you request explicitly are never held back. See Retain.
Your organization's effective rates are displayed on the Billing page, along with a per-operation cost breakdown for the last 30 days. Hover any dollar amount to see the full sub-cent precision.
How Charges Are Calculated
Tokens are counted with the o200k_base tokenizer. To estimate a charge before you send anything, count your text with o200k_base (for example, with OpenAI's tiktoken library) and multiply by the rate. The token counts on the Billing and Usage pages are the same counts you're billed on.
You're only charged for operations that succeed. Requests that fail, are rejected (including HTTP 402 for insufficient credits) or return an error such as 404 cost nothing. An asynchronous operation is charged once, when it completes, however many times it's retried.
Background processing is included. Consolidation, which turns your retained facts into observations after each retain, isn't charged separately; it's included in the retain price. Neither are the recalls that Reflect and Refresh Model run internally.
Managing your account is free. Creating, updating, reading and deleting banks; reading, listing and deleting documents; checking operation status; bank stats; listing memories; and managing API keys aren't charged, whatever the response.
Retain
Retain is charged on the content and context of each item you send. They're counted separately and added together. Nothing else is counted: not timestamps, metadata, tags, the document ID, the bank's retain mission, or the prompts Hindsight uses to extract facts.
- Re-sending a document under the same
document_idis charged only for the content that changed. Re-sending identical content costs nothing. - Appending to a document (
update_mode: "append") is charged only for the appended content. - Import (bringing documents in from an export archive, or restoring a bank) is charged on the same content and context, at the Import rate.
Recall
Recall is charged on the results it returns: each result's text plus its context. The query, chunks, entities and other response fields aren't charged.
max_tokens limits the text of the returned results, so it bounds the charge, with two exceptions:
- If no single result fits within
max_tokens, the best match is returned on its own at its full length, and charged at that length. - Each result's
contextis charged in addition to its text. If you retain without context, results have none.
Reflect and Refresh Model
Each successful call is charged the flat per-call rate, regardless of how long the answer is or how many memories it reads.
Retrieve Model
Retrieve Model is charged on the tokens of the mental model or knowledge base page content returned.
File Conversion (Iris)
Iris is charged on the tokens of the extracted text. The extracted text is then retained and charged as Retain.
Editing Memories
Editing a memory's text, or reverting an edit, is charged at the Retain rate on the text that's re-embedded. Invalidating a memory is free.
Storage
Memories stored in your banks are free for the first 30 days after they're retained. After that, you're charged $0.25 per million tokens stored per month — only the portion of your memories older than 30 days counts toward this. Your memories are never deleted.
Storage counts the content you retained (content plus context). Observations and other memories derived from it don't add to it. Deleting a document, or a whole bank, stops its storage charge from the next daily charge.
Storage charges update once per day, so the Storage tile and your balance reflect a new day's charge after the daily refresh.
The Storage tile on the Billing page shows the total tokens currently counting toward storage.
Standard vs Enhanced File Extraction
When you upload a document, you can choose between two extraction methods:
- Standard — Uses Markitdown to extract text. There is no additional charge beyond the normal Retain cost when the extracted text is stored as memories.
- Enhanced (Iris) — Uses an AI-powered vision pipeline for higher-quality extraction from scanned documents, complex layouts, and images. Iris extraction incurs an additional per-token charge based on the number of tokens in the extracted text. This charge appears as the File Conversion (Iris) line item on your billing page.
If your credit balance is exhausted, Enhanced extraction requests will be rejected with an HTTP 402 error. Standard extraction is always available regardless of credit balance.
When your credit balance is exhausted, API operations will return an HTTP 402 error until credits are added.
Transferring Credits
If you own multiple organizations, you can move credits from one to another directly from the Billing page.
Only Owners can transfer credits. You must be an owner of both the source and destination organization.
How to Transfer Credits
- Navigate to the Billing page for the organization you want to transfer credits from
- Click Transfer Credits in the Credits Balance card
- Select the destination organization from the dropdown
- Enter the amount to transfer
- Click Transfer — a confirmation dialog will appear showing the amount and destination
- Click Confirm Transfer to complete the transfer
The transfer is immediate and cannot be undone. Both organizations' transaction histories are updated — the source shows a transfer_out entry and the destination shows a transfer_in entry.
Credit transfers are permanent. Double-check the destination organization and amount before confirming.
Pay-As-You-Go
Organizations with self-service billing enabled can purchase credits directly from the Billing page.
Buying Credits
- Scroll to the Add Credits card
- Select a credit package ($10, $25, $50, or $100) or enter a custom amount ($5–$1,000)
- Complete payment via Stripe
If you have a saved payment method, you can choose to pay with it directly or proceed to Stripe Checkout. 1 credit = $1 USD.
Accepted Payment Methods
At Stripe Checkout you can pay with:
- Credit and debit cards — Visa, Mastercard, American Express, and other major networks. Cards can be saved to your account for future purchases and auto-recharge.
- Alipay (支付宝) — available for customers in mainland China. Alipay is single-use per purchase and is not saved to your account.
The payment options Stripe surfaces at checkout depend on your customer locale and the merchant configuration; expect to see the rails most relevant to your region.
Alipay payments are single-use and cannot be saved for future automatic charges. If you want to enable auto-recharge, you'll need to add a card as well — you can do this from the Payment Methods card at any time after an Alipay purchase.
Payment Methods
The Payment Methods card displays your saved credit and debit cards. From here you can:
- Add a card — Click the + button to save a new card via Stripe
- Set a default — Click the star icon on a card to make it the default for purchases and auto-recharge
- Remove a card — Click the trash icon to delete a saved card
Cards are saved automatically when you make a card purchase through Stripe Checkout. Alipay purchases do not create a saved payment method — Alipay sessions are authorized for a single charge by design.
Auto-Recharge
Auto-recharge automatically tops up your credit balance when it falls below a threshold, so your API operations are never interrupted.
- Scroll to the Auto-Recharge card
- Click Configure (or the settings icon)
- Set the following:
- Threshold — The balance amount that triggers a recharge
- Recharge amount — How many credits to add when triggered
- Daily spend cap — The most auto-recharge can charge in one UTC day. Once it's reached, further recharges are blocked until the next day
- Toggle auto-recharge on
Auto-recharge requires a saved card payment method with one set as the default. Alipay cannot be used for auto-recharge — Alipay authorizations are single-use and cannot be charged off-session. A cooldown period (default: 60 minutes) prevents rapid consecutive charges.
Auto-recharge does not guarantee uninterrupted API access. If the charge fails (e.g., due to an expired card, insufficient funds, or a payment decline), credits will not be added to your account. If your balance reaches zero, API operations will be blocked with an HTTP 402 error until credits are added manually.
Invoices
Viewing Invoices
The Invoices section shows your billing history:
| Column | Description |
|---|---|
| Invoice # | Unique identifier |
| Date | Invoice creation date |
| Amount | Total amount |
| Status | Draft, Sent, Paid, or Void |
Use the pagination controls to navigate through older invoices.
Invoice Statuses
| Status | Description |
|---|---|
| Draft | Invoice created but not finalized |
| Sent | Invoice sent to billing email |
| Paid | Payment received |
| Void | Invoice canceled |
Downloading Invoices
- Find the invoice in the list
- Click the Download PDF icon
- A PDF invoice will be downloaded
Requesting Invoice Email
- Find the invoice in the list
- Click the Send Email icon
- Invoice will be sent to your billing email
Billing Information
Updating Billing Details
- Scroll to the Billing Information section
- Update the following fields:
- Billing Name
- Billing Email
- Address (optional)
- Click Save
Billing information appears on your invoices.
Enterprise Billing
For customers on an annual contract, Hindsight Cloud supports enterprise billing with:
- Annual contract, invoiced with Net 30 payment terms
- Dedicated support
- Volume discounts
Enterprise customers receive invoices created by the Hindsight team and can have credits applied before payment clears.
Contact contact@vectorize.io for billing inquiries.