Credits and the rate card
A credit is the one unit of usage in Records Labs. Every plan includes a monthly pool, and every AI action draws from it. There is one balance per organization; members, agents, channels, and API keys all draw from the same pool.
How an action is priced
Most actions are measured: Records Labs records the real model cost of the work, converts it to credits, and rounds up to the nearest 0.1 credit. A deeper answer that reads more sources or uses a larger model costs more; a short one costs less. Answers and skill runs have a floor, a minimum they never go below. Most answers cost exactly their floor.
Some work is priced flat, per page, per minute, or per file, so you can predict it before you start.
Three rules protect you:
A request that fails before any model work starts is free. If it fails after the models have run, only the measured work is charged.
Every answer reserves credits before it starts and settles to the measured amount when it finishes. Unused credits come back immediately.
Keeping knowledge fresh is not metered: website crawls and re-crawls, connector syncing, indexing, and the enrichment of digital text are included in your plan. Reading files is metered: PDF pages, images, audio, and video use credits at the rates below, wherever the file came from. Live data on a published app refreshes for free unless its data feed runs a skill that uses AI or web search. Those refreshes use credits for the measured work, with no minimum, and show as Published apps in your usage.
The rate card
Action | Credits | Notes |
|---|---|---|
AI answer | 1 and up | Most answers cost 1. Deeper work scales with what it measures. |
Customer-facing answer | 3 and up | The minimum for an answer sent to a customer without a person in the loop: the website widget and auto-send email. Draft-mode email replies that a person edits are ordinary answers. |
Council answer | Measured | Measured across every model in the Council, so a larger Council costs more. Available on every plan. |
Search & retrieval | 0.25 | API and MCP search, retrieve, and fetch calls with no AI answer. |
Skill runs | 1 and up | Standalone API and preview runs. Scales with the work. |
Audio & video transcription | 30 per hour | Proportional to length. |
PDF & scanned-page reading | 0.5 to 1 per page | 0.5 for most pages, text layer included. 1 for complex pages read by our premium reader. On a mixed document, each page bills its own rate. |
Video visual analysis | 5 per started 30 minutes | Opt-in. Priced by length so long videos are fully understood. |
Image & media reading | 1 per file | Photos, diagrams, and screenshots read by AI. |
AI writing help | 0.25 per action | Rewrite or write-with-AI in the note editor. |
AI-drafted notes | 1 per draft | Generated from a content gap or synthesis. |
Meeting notetaker | 1.25 per minute | Recording, transcript, and summary included. |
Background corpus-analysis reports bill like answers, measured from the work done and charged once when the job completes. Cancelled or failed jobs are not billed.
The rate card is versioned. When a rate changes, history is never re-priced; each charge records the version it was billed under. The current card is shown under Settings → Plan & Usage → Usage in the What credits cost card.
Order of use
Credits come out of your monthly pool first, then out of purchased extra credits. There is no postpaid overage on any self-serve plan: when both are gone, AI work pauses until you buy credits or auto-reload buys them for you. See Extra credits and auto-reload.
Where to see usage
Settings → Plan & Usage (owners, admins, and roles with "Billing & plan"):
On the Overview tab, Credits this cycle shows the pool, extra credits, and what is left.
On the Usage tab, Usage this period shows what each kind of work cost this month.
Top consumers (Usage tab) lists the five members who spent the most this month, your early warning before you run low.
What credits cost (Usage tab) is the live rate card.
Analytics → My usage is open to every member and shows their own usage. Admins also see organization-wide usage and cost views in Analytics.
Alerts
Admins get a notification when the pool reaches the percentages set under Usage warnings (75% and 90% by default). The 90% and 100% alerts always send. If your monthly plan credits stay used up and no one turns on auto-reload or buys credits, admins get a reminder 24 hours after the 100% alert and then every three days until the pool resets. The reminders come even while older extra credits are still being used.