Understand Your Credits
Credits are the currency that pays for everything your AI employees do.
TL;DR
Every time an employee thinks, calls a paid tool, generates media, or takes a voice or meeting minute, it spends credits from your workspace balance. At current monthly USD list prices, 1,000 credits costs $2.50 on Starter, Builder, Founder and Leader. Discounts, existing contracts and top-ups can have different prices. Watch your balance in the top-right corner of your workspace, or see the full breakdown under Settings, Subscription, Credit Usage.
New Free accounts include 2,000 one-time credits. They do not refill each month. Existing accounts keep their current allowance and balance. Promotional and referral credits are shown separately in your credit history.
What Is a Credit?
A credit is the billing unit for all AI work on the platform. Instead of tracking dollars for every model call and tool invocation separately, everything is measured in credits: your workspace has a credit balance, employees spend credits when they work, and you upgrade your plan when you need more.
The current monthly USD plans share the same list price per credit. Larger plans provide a larger allowance and additional features. Credits spent on an action depend on the model, its input and output usage, any available prompt-cache savings, and paid tools used. One reply can involve several model or tool calls.
Credits belong to your workspace, not to you personally. If you own more than one organization, each one has its own separate plan and its own separate credit balance; usage in one never draws down the other.
What Costs Credits
AI Usage (the biggest factor)
Every time an employee thinks, responds, or processes information, it uses an AI model. The cost depends on the specific model and its measured usage. The tiers group model capabilities; they are not fixed billing rates.
| Tier | Best For | Approximate Cost |
|---|---|---|
| Fast | Simple lookups, formatting, drafts, quick Q&A | Often lower cost; varies by model |
| Standard | Everyday tasks, conversations, analysis | Medium |
| Advanced | Complex reasoning, long documents, nuanced work | High |
| Frontier | The most capable models available (only tier that skips the "not on the best brain" notice) | Varies by model and usage |
You choose the model tier per employee on their Profile tab, under the Brain section (Employee, Profile, Brain). Fast is great for routine work and saves credits; Advanced and Frontier are best for tasks that need deeper thinking. The exact credit charge for a given message always reflects the real cost of the model and tokens used, not a flat per-message fee.
Prompt caching and supporting AI work
Saving input for reuse can cost more on the first call; reading that saved input on later calls costs less. Both use the selected model's rates. Savings depend on how often the input is reused.
Memory updates, knowledge searches, enabled safety and quality checks, and desktop browser-agent steps can use paid AI calls too. They appear as separate usage. Quality checks are optional; several enabled checks or correction attempts can mean several charges for one response. Completed provider work can still cost credits if a later save or upload fails.
Other Credit Costs
| Activity | What Determines the Credit Charge |
|---|---|
| Voice calls | Measured speech recognition, generated speech, model work and any video/avatar service used |
| Meeting attendance | Measured meeting-provider usage and any separately billed AI processing |
| Image generation | Model, quality, size and measured text/image tokens |
| Video generation | Model, duration, resolution, aspect ratio, audio and provider-reported generation cost |
| Connected app calls | Measured charges from paid connected services; free operations add no tool charge |
| Mailbox emails sent | Measured billable outbound-email provider usage |
These are usage-based charges, not universal flat fees per message, minute or image. Any AI model work used to carry out an action is included separately in its usage breakdown.
What Does NOT Cost Credits
- Task creation and completion, workflow starts, and compute runtime while a workflow runs
- Web Search using the default search engine, file management, journal entries, sprint management
- Internal processing like database queries and cache lookups
- Using a local AI model through Ollama (zero credits for the AI thinking itself)
Where to See Your Credit Balance
The Credit Bar (always visible)
Look at the top-right corner of your workspace. You will see your current plan name, a progress bar showing how much of your monthly allowance remains, and the percentage left. This updates as your employees work; the bar turns yellow under 25% remaining and red under 10%.
Click the credit bar to open a dropdown with more detail:
- Your remaining credit count and a progress bar
- A breakdown of where credits went this billing period: AI usage, tool calls, internal AI operations, compute, tasks, voice calls, and emails sent
- Quick links to upgrade your plan or view full usage history
Per-Message Cost
Every chat message from your employee shows how many credits it cost. Click the inspect button on any message to see a detailed breakdown: input tokens, output tokens, and tool usage.
Credit Usage Page (Full Details)
For a complete picture, go to Settings (gear icon in the sidebar), Subscription, Credit Usage.
This page has five sections:
Credit Usage (current billing period)
A detailed breakdown of spending this month, including AI usage by tier and model, tool calls, internal AI operations, and voice and mailbox activity. Most of these line items reflect the model and tool costs above; compute time and task completion currently show as zero cost. Each line counts a charge once, and the Total charged row at the bottom is the sum of the lines above it, matching the total in Credit Spend Breakdown.
Credit Spend Breakdown
See the credits in this period that followed an instruction you sent to an employee and the credits from automated or system work. The card separately shows how many distinct direct instructions caused charged work; that smaller number is not a credit total or your total message count.
Direct instructions include paid work started from a human request in chat, Slack, or another human channel. Routine and proactive work still consume credits when they call paid models or tools, even when no person sends a new request. The Automation & system figure is the reconciled remainder. If an older ledger row has no recoverable original trigger, the page discloses its amount and includes it in that remainder only so the charged total stays exact.
The chart below is the same ledger grouped by the operation that incurred each charge: AI work, connected tools, Officer checks, knowledge and training, voice, meetings, video, emails, or any other charged action. It lists the exact credits and recorded operations for every charged group. This is a different question from how the work started, so the two totals are deliberately kept separate.
Credit Runway
A single chart shows your remaining balance from the start of this month to today, then a dashed projection through the reset date. It names your recorded daily burn rate and tells you whether credits are on track to last or when they may run out.
Usage History
A month-by-month breakdown of your credit consumption and work activity.
Credit Grants
A table of all bonus credits you have received: promo codes, referral rewards, and admin grants. Each row shows the date, source, and amount.
How to Earn FREE Credits
Referral Program
Invite people to the platform and you both earn credits. Here is how:
- Go to Settings (gear icon), Subscription, Get Free Credits
- You will see your personal referral link and stats
- Share your link by copying it, or send an email invite directly from the page
- When someone signs up using your link, you both receive 1,000 bonus credits
- You can earn referral credits for up to 10 successful signups
The page shows your invites sent, how many were accepted, and total credits earned.
Free Plan
New signups land on a plan picker with no credits until they choose a plan. Picking the Free plan costs nothing and no card is required: it grants the current lifetime allowance, which never resets. See Technical Limits in your workspace for the live amount before deciding on a paid plan.
Promo Codes
From time to time, promotional codes are available that grant bonus credits. Redeem one from the same Get Free Credits page at any time, not only at signup. Redeemed codes show up in your Credit Grants table.
Credit Alerts
You do not need to watch your balance constantly. The platform notifies you automatically:
| Threshold | What Happens |
|---|---|
| 25% remaining | You receive an in-app notification that credits are getting low |
| 15% remaining | A more urgent notification warns you credits are critically low |
| 0% remaining | A final notification tells you credits are exhausted |
Each alert only re-fires if things get worse, so you will not be spammed with repeats at the same level.
What Happens When Credits Run Out
When your balance hits zero:
- The chat composer locks and shows a modal telling you you've used all your credits, with a button to upgrade your plan
- Any active streaming response stops
- Voice calls end automatically
- Scheduled and autonomous work is skipped until credits are available again; nothing errors out, the schedule stays active and simply resumes on its own
To resume: upgrade your plan, or wait for your next billing period when your monthly credits refresh. Everything comes back online automatically once credits are available.
Tips & Tricks
- Use Fast-tier models for routine work. Quick lookups, formatting, and simple drafts cost a fraction of what Advanced or Frontier models cost. Switch employees to Fast when they do not need deep reasoning
- Check the credit bar dropdown regularly. It shows exactly where your credits are going this month, broken down by category
- Set up referrals early. Each successful referral earns you 1,000 credits for free, and your friend gets 1,000 too
- Redeem promo codes as soon as you get one. They can be applied any time from the Get Free Credits page, not just at signup
Frequently Asked Questions
Q: What happens when credits run out? A: All AI work pauses. Chat locks with an upgrade prompt, voice calls end, and scheduled work waits until credits return. You are warned at 25% and 15% remaining. Upgrade your plan or wait for your next billing cycle to resume.
Q: Can I see how many credits a single message costs? A: Yes. Every message shows its credit cost. Click the inspect button for a full breakdown by model and token type (input, output, reasoning).
Q: Do unused credits roll over to the next month? A: No. Monthly plan credits reset at the start of each billing period. Free plan credits are the exception: they are a one-time lifetime grant that never resets, but also never refills.
Q: Are local AI models (Ollama) free? A: Yes. If an employee uses an Ollama model running on your own machine, zero credits are charged for the AI thinking.
Q: Can I set a credit budget per employee? A: Not currently. Spending limits are set at the workspace level (Technical, Rate Limits), not per employee.
Q: What is the difference between Fast, Standard, Advanced, and Frontier? A: These are model tiers. Fast models focus on quick responses, Standard models suit everyday work, and Advanced and Frontier models offer stronger capabilities for demanding tasks. Pricing belongs to each model, so a tier is not a fixed credit charge.
Q: How are credits tracked if I own multiple organizations? A: Credits are tracked per organization, not per user. If you own two organizations, each one has its own plan and its own separate credit balance; using one does not draw down the other.