Credits
Credits are the unit of work in MarketingSecrets.ai. Every AI call — chat, routine, Sprint App sync — consumes credits. Your plan includes a monthly allowance, and routines use credits passively in the background.
What drives the cost of a request
There's no fixed price per message. What you pay for a single request comes down to three things:
- How long the conversation is. This is the big one. Every request re-reads the conversation so far, so a reply late in a long thread costs meaningfully more than the same question asked in a fresh chat — on the same model.
- How much work it takes. A one-line answer is cheap. Searching the web, reading a document, calling a connected app, then writing it all up is several pieces of work in one request.
- Which model is running. Models differ in cost, and the picker shows each one's cost relative to the default.
Because of #1, no honest per-message number exists — the same model on the same question can cost very differently depending on where you are in a conversation. So the picker shows relative cost ($ → $$$$$), and your Credits page shows exactly what each request actually cost.
Starting a fresh chat for a new topic is the single most effective way to keep costs down.
Reading the model picker
Two different things are marked, with two different icons:
- 🧠 Brains next to a group = its capability tier (Fast / Balanced / Premium / Frontier)
- The
$cost badge on a model = how expensive it is relative to the default model
They don't track each other — some Frontier-grade models are inexpensive, and some Premium ones aren't.
Hover any badge to see what it means in words.
| Badge | What it means |
|---|---|
$ | Lowest cost — Cheaper than the default model |
$$ | Low cost — About the same as the default |
$$$ | Moderate cost — A few times the default |
$$$$ | High cost — Many times the default |
$$$$$ | Highest cost — The most expensive tier, best saved for genuinely hard problems |
New chats start on your default model, which is fast and inexpensive. If you never touch the picker, you're on a low-cost model.
Every model has a per-turn ceiling so a single message can't run away. And you aren't charged for a request that failed. If a message errors out, times out, crashes part-way through, or finishes without producing anything usable, it costs you nothing. Retry as many times as you need: every attempt that fails is free, and you only pay once something comes back. Failed attempts still appear on your Credits page, at 0 credits, so you can see they happened. Tools that error out are free the same way: you're only charged for the ones that came back with something.
What each feature costs
These are flat rates, added on top of the chat reply that triggered them. They're the same regardless of which model you're on.
| Feature | Credits |
|---|---|
| Save a story to your Attractive Character | 1 |
| Create a routine or agent | 5 |
| Find past chats and files (semantic search) | 10 |
| Generate an image | 10 |
| Web search | 15 |
| Deep research report | 90 |
| Generate a video | 250 |
Everything else your Chief of Staff does — remembering facts, searching Russell's training, searching your own files, creating and updating documents, saving a chat to a Project, running skills, emailing yourself a summary, reading from your connected apps — carries no separate charge beyond the chat reply.
Deep research (90) and video (250) are the only two that visibly move a balance — many times what a typical chat reply costs. Both ask for your confirmation first (see below). Everything else is small enough that you won't notice it.
You can see this same list any time inside the app: tap the ? icon next to the chat composer.
Confirmation before an expensive request
Two protections stop a surprise charge:
- Deep research and video generation always ask first. They never fire automatically. You get a card showing the estimated cost, and nothing is charged until you approve it. If you cancel, you pay nothing.
- Any unusually expensive message asks first. If a single message is going to cost far more than a normal turn — usually a premium model plus a very large pasted document — you'll get a confirmation card showing the estimate before it runs. The same card appears if one message would eat a large share of your remaining balance. You can approve it, switch to a cheaper model, or cancel.
Because these gates fire before the work happens, an approved request is the only one you pay for.
View usage
Open the Credits page (account menu at the bottom of the sidebar) to see:
- Current Balance at the top, split into plan, bonus, and purchased credits
- Usage This Month, broken out by action (last 30 days)
- Every charge itemized, including which model ran and which tools it used
- Top Up Credits packs when you need more — any member can buy, and the credits land in the workspace's shared pool
If a specific charge ever looks wrong, that page is the record — and support can review any individual session for you.
Brain Setup Score
Each setup task you complete makes your Chief of Staff smarter — and earns bonus credits. Finish them all for 1,150 bonus credits. Track progress on the Chief of Staff page.
| Task | Bonus |
|---|---|
| Finish setup | +250 credits |
| First conversation | +50 credits |
| Connect Gmail | +150 credits |
| Connect Calendar | +100 credits |
| Connect Stripe | +100 credits |
| Connect Instagram | +100 credits |
| Upload a document | +100 credits |
| Import ChatGPT memories | +200 credits |
| First routine runs | +100 credits |
A bonus counts however the task gets done. If your Chief of Staff builds your Attractive Character (+100 credits) or a Customer Avatar (+75 credits) for you in a conversation, or ChatGPT or Claude creates the character through your Brain connection, you earn the same bonus as when you make it yourself. Each bonus is paid once.
Passive burn warnings
Routines can use credits while you sleep. To prevent surprises:
- Pause a routine if you're not using its output — one click on the Routines page
- Routines stop at zero — scheduled runs are skipped automatically when your balance runs out, so you never go negative in the background
- Keep routines on the default model — routines run on the cost-efficient default unless you ask for a premium one. The routine editor shows each model's relative cost.
- A runaway routine auto-pauses — if one routine uses more than 300 credits in 24 hours, we pause it and tell you, rather than letting it drain your balance
Every routine run is logged with its credit cost — audit anytime from your Chief of Staff Actions page.
Auto Top-Off
Never want to run out mid-task? Turn on Auto Top-Off on the Credits page and add a payment method — when your balance drops below 200 credits, it recharges automatically in the background, and refills are discounted versus one-time buys. You pick the refill amount; workspace owners manage it, and you can turn it off with one click.
Earn more credits
- Refer a friend — +1,000 credits for you and +1,000 for them when they finish setup
- Work through Sprints — +50 for your first Sprint video, +200 for completing a Sprint, +100 for installing your first Sprint App
- Show up daily — +50 credits for your first login each day (resets at midnight UTC, so the exact local time it becomes available varies), plus +200 credits for a 7-day streak and +500 for a 30-day streak
- Build your team — +100 credits when you invite your first team member
Do credits roll over?
Your balance is three buckets: plan credits reset each billing period (unused plan credits do not roll over), while bonus and purchased credits never expire. Plan credits are spent first, then bonus, then purchased. The full rollover, reset, and refund answers live in the Plans, Credits & Refunds FAQ.
Next
- Plans, Credits & Refunds FAQ — rollover, resets, refunds, plan pricing
- Routines — see which ones consume the most credits
- Billing & Cancellation — manage or cancel your plan