How to Monitor AI Agent Usage and Costs

Understanding how your AI agents consume credits is essential for managing costs and getting the most value from your EZClaws subscription. The EZClaws dashboard provides real-time visibility into every aspect of your usage — from total credit consumption to per-agent breakdowns to individual token costs.

In this guide, you will learn how to navigate the usage monitoring features, understand the credit system, set up alerts, and use the data to make informed decisions about your AI agent operations.

Prerequisites

To follow this guide, you need:

• An EZClaws account — Sign up at ezclaws.com if you have not already.
• At least one deployed agent — You need some usage data to monitor. Deploy an agent with our deployment guide if needed.
• Basic familiarity with the EZClaws dashboard — You should know how to log in and navigate to your agents.

Step 1: Access the Billing Dashboard

Log in to your EZClaws account and navigate to the billing section at /app/billing. This is your central hub for all usage and cost information.

The billing dashboard displays:

• Current Plan — Your subscription tier and its included credit allocation.
• Credit Usage — A visual progress bar showing credits used vs. available.
• Billing Cycle — When your current cycle started and when it ends.
• Usage History — A timeline of credit consumption over the current and past cycles.

Example dashboard view:

Plan: Pro
Credits: 4,250 / 10,000 cents used (42.5%)
Cycle: Feb 1, 2026 — Feb 28, 2026
Estimated end-of-cycle usage: 9,200 cents

The dashboard updates in real time using Convex's reactive data layer, so you see credit changes as they happen.

Step 2: Understand the Credit System

EZClaws uses a credit-based system measured in cents. Here is how it works:

Credit Allocation

Each billing cycle (typically monthly), your plan allocates a credit budget:

• Starter Plan — Included credits suitable for light usage.
• Pro Plan — Higher allocation for active agents.
• Business Plan — Generous allocation for production workloads.

Visit /pricing for current plan details and credit allocations.

Credit Consumption

Credits are consumed when your agents make API calls to model providers. The cost depends on:

1. Model — Larger, more capable models cost more per token.
2. Input tokens — The text sent to the model (user message, system prompt, conversation history, memory context).
3. Output tokens — The text generated by the model (the agent's response).

Cost per Model

Here is an approximate cost breakdown for popular models:

Model              | Input (per 1M tokens) | Output (per 1M tokens)
-------------------|-----------------------|-----------------------
GPT-4o             | $2.50                 | $10.00
GPT-4o-mini        | $0.15                 | $0.60
Claude 3.5 Sonnet  | $3.00                 | $15.00
Claude 3 Haiku     | $0.25                 | $1.25
Gemini Pro         | $1.25                 | $5.00

Note: Prices are approximate and may change. Check your model provider's pricing page for current rates.

Example Cost Calculation

A typical conversation might involve:

Input: 500 tokens (user message + system prompt + context)
Output: 200 tokens (agent response)

Using GPT-4o:
Input cost: 500 / 1,000,000 * $2.50 = $0.00125
Output cost: 200 / 1,000,000 * $10.00 = $0.002
Total: $0.00325 (about 0.3 cents)

Using GPT-4o-mini:
Input cost: 500 / 1,000,000 * $0.15 = $0.000075
Output cost: 200 / 1,000,000 * $0.60 = $0.00012
Total: $0.000195 (about 0.02 cents)

This means a GPT-4o conversation is roughly 16 times more expensive than GPT-4o-mini. Understanding this helps you choose the right model for each task.

Step 3: View Per-Agent Usage

From the billing dashboard, you can drill down into individual agent usage:

1. Navigate to your agent's detail page at /app/agents/[id].
2. Look for the Usage section.
3. You will see:
   • Total credits consumed by this agent in the current cycle.
   • A breakdown by day or hour.
   • The models used and their respective costs.
   • Number of requests processed.

Comparing Agent Usage

If you have multiple agents, compare their usage to identify which ones consume the most resources:

Agent              | Credits Used | Requests | Avg Cost/Request
-------------------|-------------|----------|------------------
Support Bot        | 2,100 cents | 1,400    | 1.5 cents
Research Assistant  | 1,800 cents | 200      | 9.0 cents
Telegram Bot       | 350 cents   | 350      | 1.0 cents

In this example, the Research Assistant processes fewer requests but each request is more expensive because research tasks involve more tokens (web browsing, long responses, multi-step reasoning).

Step 4: Analyze Usage Records

EZClaws stores detailed usage records for every API call. Each record includes:

• Timestamp — When the request was made.
• Agent ID — Which agent made the request.
• Model — Which model was used.
• Provider — The model provider (OpenAI, Anthropic, etc.).
• Input Tokens — Number of tokens in the request.
• Output Tokens — Number of tokens in the response.
• Cost — The credit cost in cents.

You can use this data to:

• Identify expensive conversations or requests.
• Spot usage spikes and correlate them with specific events.
• Determine if you are using the right model for each task.
• Plan for future credit needs based on historical patterns.

Step 5: Set Up Usage Alerts

Proactive monitoring is better than reactive. Set up alerts to notify you before issues occur.

Navigate to your settings at /app/settings and configure notifications:

Low Credit Alert

Set a threshold for low credit notifications:

Alert when credits fall below: 20% of total allocation
Notification method: Email

This gives you advance warning to purchase additional credits or optimize usage before running out.

Weekly Usage Summary

Enable weekly email summaries that include:

• Total credits consumed during the week.
• Per-agent breakdown.
• Comparison with the previous week.
• Projected end-of-cycle usage.

Unusual Usage Alert

Configure alerts for abnormal usage patterns:

Alert when daily usage exceeds: 200% of 7-day average
Notification method: Email

This catches situations where an agent suddenly consumes more credits than expected, which could indicate a configuration issue or unexpected traffic spike.

Step 6: Optimize Based on Data

Use your usage data to make informed optimization decisions:

Identify Expensive Operations

Look for patterns in your usage records:

• Are web browsing tasks consuming excessive tokens?
• Is conversation history growing too large?
• Are there redundant API calls?

Right-Size Your Models

If an agent handles a mix of simple and complex tasks, consider using different models:

• Simple queries (FAQ, basic info) — Use GPT-4o-mini or Claude Haiku (cheapest).
• Complex tasks (research, code, analysis) — Use GPT-4o or Claude Sonnet (most capable).

See our model provider configuration guide for how to set this up.

Reduce Token Usage

• Keep system prompts concise — every word in the system prompt is sent with every request.
• Configure memory limits — cap the amount of conversation history included in each request. See memory guide.
• Use shorter responses where appropriate — instruct the agent to be concise.

For a comprehensive cost optimization guide, see How to Reduce AI Agent API Costs.

Plan Your Budget

Use historical data to project future costs:

Current daily usage: ~150 cents
Days remaining in cycle: 15
Projected remaining usage: 2,250 cents
Current remaining credits: 5,750 cents
Status: Comfortable — no action needed

If projections show you will exceed your allocation, either:

1. Optimize usage (reduce unnecessary calls, switch to cheaper models).
2. Purchase additional credits.
3. Upgrade to a higher plan at /pricing.

Troubleshooting

Dashboard shows no usage data

1. Check agent activity — Ensure your agent has actually processed some requests.
2. Verify agent status — A stopped agent does not generate usage data.
3. Wait a moment — There can be a brief delay between API calls and dashboard updates.
4. Refresh the page — Though the dashboard is real-time, a refresh can help if there is a connection issue.

Usage seems higher than expected

1. Check for long conversations — Extended conversations accumulate tokens across all messages in the history.
2. Review model selection — You might be using a more expensive model than needed.
3. Check system prompt length — A verbose system prompt adds tokens to every request.
4. Look for automated traffic — If your agent is connected to a public channel, unexpected users may be generating requests.
5. Review skill token usage — Some skills (web browsing, code execution) generate additional API calls.

Cannot access billing page

1. Check authentication — Ensure you are logged in.
2. Clear browser cache — Try in an incognito window.
3. Check account status — Ensure your account is in good standing.

Usage data does not match model provider's dashboard

Small discrepancies are normal due to:

• Token counting differences between providers.
• Timing differences in reporting.
• Rounding in cost calculations.

If the discrepancy is significant, contact EZClaws support with specific examples.

Understanding Your Usage Cycle

EZClaws billing cycles work as follows:

1. Cycle Start — Credits are allocated based on your subscription plan.
2. During the Cycle — Credits are consumed as agents make API calls.
3. Low Credits — Notifications alert you when credits are running low.
4. Cycle End — Unused credits do not roll over. A new allocation begins with the next cycle.

One-Time Credit Purchases

If you need additional credits mid-cycle, you can purchase one-time credit packs. These are added to your current cycle balance immediately and are available until the cycle ends.

Plan Changes

Upgrading your plan mid-cycle immediately adjusts your credit allocation. The change is prorated based on the remaining days in the cycle.

Summary

Monitoring your AI agent usage is straightforward with the EZClaws real-time dashboard. The billing section at /app/billing gives you complete visibility into credit consumption, per-agent breakdowns, and usage trends. Setting up alerts ensures you are never caught off guard by credit exhaustion.

The key to effective monitoring is regular review combined with proactive alerts. Check your dashboard weekly, analyze usage patterns, and optimize based on data. This approach keeps your costs predictable while ensuring your agents have the credits they need to operate effectively.

For detailed cost optimization strategies, continue to our cost reduction guide. For more on managing your EZClaws setup, explore our blog and other guides.