Credits & Billing
The One Agents uses a credit-based model for AI compute. Credits are the only thing you pay for beyond your monthly subscription — every non-AI step in every agent is free.
How Credits Work
One Agent Credit equals approximately $0.001 retail. Credits abstract away the complexity of token counting and model pricing. Instead of tracking "8 million GPT-4o mini tokens per month," you track a single credit balance.
Each AI Reason step in an agent consumes a fixed number of credits based on the model tier selected. The cost is the same regardless of the actual token count — you pay for the tier, not the tokens.
Non-AI steps cost zero credits:
| Step Type | Credits |
|---|---|
| AI Reason — Quick | 5 |
| AI Reason — Standard | 20 |
| AI Reason — Premium | 100 |
| AI Reason — Expert | 200 |
| Product Action | 0 |
| Notification | 0 |
| Approval Gate | 0 |
| Branch | 0 |
| Wait | 0 |
Model Tiers
Each AI Reason step requires choosing a model tier. Higher tiers use more powerful models and cost more credits.
| Tier | Model | Token Budget | Credits | Wholesale Cost | Retail Cost | Best For |
|---|---|---|---|---|---|---|
| Quick | GPT-4o mini | ~2K tokens | 5 | ~$0.0006 | $0.005 | Simple classification, routing, binary decisions |
| Standard | GPT-4o mini | ~8K tokens | 20 | ~$0.0026 | $0.020 | Triage, draft responses, moderate analysis |
| Premium | GPT-4o | ~8K tokens | 100 | ~$0.045 | $0.100 | Complex analysis, QBR narratives, detailed summaries |
| Expert | Claude Sonnet | ~8K tokens | 200 | ~$0.063 | $0.200 | Highest-quality reasoning, compliance analysis, nuanced judgment |
Rule of thumb: Default to Standard for most automation steps. Use Quick for steps that just need to pick from a list. Reserve Premium and Expert for steps where the quality of reasoning directly impacts a client relationship or security decision.
Monthly Subscription Tiers
| Tier | Price | Credits/mo | Max Custom Agents | Best For |
|---|---|---|---|---|
| Starter | $99/mo | 20,000 | 5 active | Small MSPs (≤20 clients), getting started with automation |
| Growth | $179/mo | 100,000 | 25 active | Most MSPs (20–100 clients), active automation program |
| Scale | $299/mo | 300,000 | Unlimited | Large MSPs (100+ clients), aggressive automation |
All tiers include:
- All four system agents (always on, don't count against custom agent limit)
- Full template library (clone and customize any template)
- Client Intelligence sidebar across all products
- Event Catalog access
- Complete execution audit trail (90-day retention)
Credits expire monthly — unused credits do not roll over. This simplifies billing and encourages consistent use.
Typical Credit Consumption
Here's what credit consumption looks like for a typical 50-client MSP using the four system agents plus a few custom agents:
| Agent | Activity | Credits/Step | Weekly Runs | Weekly Credits |
|---|---|---|---|---|
| Dispatch Agent | ~100 tickets/week | 20 (Standard) | 100 | 2,000 |
| Guardian Agent | ~30 alerts/week | 20 (Standard) | 30 | 600 |
| Advisor Agent | Daily digest | 5 (Quick) | 7 | 35 |
| Context Builder | No AI steps | 0 | — | 0 |
| Custom: P1 Escalation | SMS only, no AI | 0 | ~5 | 0 |
| Custom: Security Response | 1× Premium/critical alert | 100 (Premium) | ~2 | 200 |
Weekly total: ~2,835 credits Monthly total: ~11,340 credits
At Growth tier (100,000 credits/mo), this MSP uses roughly 11% of their allocation — substantial headroom for additional custom agents and growth.
Monitoring Usage
Navigate to Credits in The One Agents to see:
Current Month
- Credits used — total consumed this month
- Credits remaining — allocation minus consumed
- Usage by agent — which agents are consuming the most credits
- Usage by model — breakdown across Quick, Standard, Premium, Expert
12-Month History
Month-over-month trend chart showing credit consumption. Useful for forecasting and identifying usage spikes from new agents or high-alert periods.
80% Alert
When your organization reaches 80% of your monthly allocation, The One Agents sends an automatic notification to your Hub admins. This gives you time to:
- Review which agents are consuming the most credits
- Pause low-priority agents for the rest of the month
- Purchase a top-up pack to cover the remainder
Top-Up Packs
If you exhaust your monthly allocation before the end of the month, top-up packs let you add credits immediately without upgrading your subscription tier.
| Pack | Credits | Price | Cost per Credit |
|---|---|---|---|
| Small | 10,000 | $19.99 | $0.002 |
| Medium | 50,000 | $89.99 | $0.0018 |
| Large | 100,000 | $169.99 | $0.0017 |
Top-up credits are consumed before your monthly allocation. They do not expire at month end — they roll over until fully consumed.
Credit Depletion Behavior
When your monthly allocation reaches zero and no top-up packs are available:
- AI Reason steps begin failing with a
credit_insufficienterror - Non-AI steps (Product Action, Notification, Branch, Wait, Approval Gate) continue running unaffected
- Executions with failed AI steps are logged with the error — the agent continues if
on_error: continue, fails ifon_error: fail - No new AI reasoning occurs until the next month's allocation resets or a top-up pack is purchased
System agents are subject to the same credit constraints. If your allocation is depleted, Dispatch Agent will not classify tickets and Guardian Agent will not enrich alerts until credits are restored. Non-AI steps in those agents (ticket creation, notifications) will still execute.
Credit Allocation Best Practices
Audit before upgrading. Before purchasing a top-up or upgrading, review your usage-by-agent breakdown. Often, a single misconfigured agent firing on high-volume events is responsible for the overage.
Right-size model tiers. Standard (20 credits) handles most MSP classification and triage scenarios well. If you're using Premium or Expert for steps that don't require that level of reasoning, downgrade them.
Use trigger filters aggressively. An agent that fires on every psa:ticket.updated event (including system-generated updates) will consume far more credits than one filtered to only status eq in_progress. Check your execution counts in the Executions view.
Scope agents to relevant clients. If an agent only makes sense for clients with Defend enrolled, scope it to those client IDs. This prevents unnecessary credit consumption for clients where the automation doesn't apply.