HelixDB Docs

Guide

Reading documents into loads and answering questions use AI models, so they cost money; reading tables doesn’t. Each has a monthly cap per workspace, the same on every plan. Select your name at the foot of the sidebar: Usage, in the account menu, shows what the workspace has spent this month.

The account menu's Usage: this month's total, then Loads of $50.00, API sources within it of $20.00, and Questions of $15.00, each with a bar

Row	What it counts	Cap each month
This month	Everything below, once
Loads	Reading documents, and API sources’ pieces, into your ontologies’ loads	$50
API sources	Part of Loads: extracting API sources’ pieces that no one started a load for	$20
Questions	Answering questions in Ask and through the Helix MCP	$15

Each row’s bar fills as the month’s spend nears its cap. Point at a row for what it counts; point at the total for the day the caps start again, the first of each month (UTC). Everyone in the workspace sees it. Spend on the workspace’s own model key counts toward the caps too.

When a cap is reached

Cap	What happens
A load: $10	The load stops Partial, with its tables written. The documents it didn’t reach are read into it once it’s current and the month has budget.
Loads: $50 a month	New loads are refused, and new documents aren’t read, until next month. Work no person started, such as refreshes and reading documents as they arrive, stops at $40, which leaves the last $10 for loads you start.
API sources: $5 per source, $20 a month	Extraction no person started stops for the month: new pieces wait for next month, and ontologies keep the pieces they have.
A question: $0.50	The question stops at its limit. Ask something narrower.
Questions: $15 a month	A question waits until the month’s budget can hold it, or you cancel it.

A refresh that runs with the month spent still writes your tables, and reuses what earlier loads read from documents at no cost. Drafting an ontology isn’t in Usage: each generation attempt stops at $5.

Next steps

Limits

Every limit in one place.

AI models

Where the workspace’s AI runs.