Cost & Audit

This article covers Monstrum’s Data Center features, including three modules: Cost Operations, Security Operations, and Audit Logs.


Overview

Monstrum records and tracks every operation performed by a Bot. The Data Center is your central hub for understanding platform activity, and it includes three modules:

ModulePurpose
Cost OperationsToken consumption, cost statistics, budget management
Security OperationsPermission check statistics, security events
Audit LogsFull-chain operation logs

Click Data Center in the left navigation bar to enter.


Cost Operations

Cost Overview

The Cost Operations page displays:

  • Monthly Cost: Total LLM call cost (USD) across all Bots for the current month
  • Total Cost: Cumulative historical cost
  • Token Statistics: Prompt Token + Completion Token consumption

Charts show token usage over time, with support for filtering by time range:

Time RangeDescription
Last 1 hourReal-time monitoring
Last 24 hoursIntraday trends
Last 7 daysWeekly trends
Last 30 daysMonthly trends

Per-Bot Statistics

View each Bot’s token consumption and cost distribution to quickly identify the highest-consuming Bots.

Budget Management

Each Bot can have a monthly budget (USD) configured in its settings:

  • Budget Warning: When token consumption reaches 80% of the budget, the platform sends a budget warning notification (triggered at most once per session)
  • Budget Enforcement: After exceeding the monthly budget, the Bot automatically stops executing new LLM requests
  • Budget Reset: Automatically resets each month

Configure the budget amount in the Settings tab of the Bot detail page.


Security Operations

Permission Check Statistics

Charts display permission check pass rates and rejection counts:

  • Permission Pass Rate: The proportion of tool calls that passed permission checks
  • Allow/Deny Distribution: Distribution of permission check results

A high rejection rate may indicate:

  • The Bot’s permission configuration needs adjustment (permissions too strict)
  • The Bot is exhibiting abnormal behavior (attempting unauthorized operations)

Task Completion

  • Task Success Rate: The proportion of tasks completed successfully
  • Success/Failure Distribution: Distribution of task execution results

Audit Logs

Operation Logs

Audit Logs record every operation performed by a Bot, forming a complete audit trail.

Each log entry contains:

FieldDescription
TimeWhen the operation occurred
Operation TypeTool call / LLM request / Permission check / Task lifecycle
BotThe Bot that performed the operation
Task IDThe associated task
PermissionPermission check result (Allow / Deny)
DetailsOperation detail summary

Operation Types

TypeDescription
Tool CallBot called an external resource tool
LLM RequestBot called an LLM for reasoning
Permission CheckPermission verification event
Task StartedTask began execution
Task CompletedTask finished execution
Task FailedTask execution failed

The log list supports multi-dimensional filtering:

  • Bot Filter: View operation logs for a specific Bot
  • Operation Type: Filter by type (Tool call / LLM request / Permission check, etc.)
  • Permission Result: View only allowed or denied records
  • Time Range: Specify start and end times

Log Details

Click a log entry to view details, including:

  • Full parameters of tool calls
  • Token consumption and model information for LLM requests
  • Dimensions and constraint values for permission checks
  • Execution duration

Export

Logs can be exported as CSV files, containing the time, operation, task ID, Bot, permission, and details fields.


Task Monitoring

Task List

In the Data Center, you can view task execution records for all Bots:

  • Task Status: Pending / Running / Completed / Failed / Cancelled
  • Associated Bot
  • Task Instructions
  • Token Consumption
  • Execution Duration

Supports filtering by status and Bot.

Task Details

Click a task to enter the detail page and view:

  • The full task instructions
  • Each step during execution (reasoning output, tool calls, execution results)
  • Total token consumption and cost
  • Checkpoint status (if applicable)

Execution Replay

The task detail page supports execution replay — a timeline view of the Bot’s complete task execution process, including each round of reasoning output and tool call results.


Execution Location

Operation logs record the execution location of tool calls:

Execution LocationDescription
cloudExecuted on the platform cloud
agent:{name}Executed locally via an Agent

This allows you to distinguish which operations were executed on the platform side and which were executed locally through an Agent in the user’s environment.


Notifications

The platform sends notifications for key events:

Notification TypeTrigger Condition
Task FailedA task execution failed
Budget WarningToken consumption reached 80% of budget
Approval RequiredA workflow step requires manual approval
Tool ChangesThe tool list of an MCP resource changed (added/removed)

Common Use Cases

Cost Optimization

  1. Identify the highest-consuming Bots in Cost Operations
  2. Check the Bot’s reasoning mode — adaptive mode consumes the most, reactive the least
  3. Check the Bot’s maximum iteration count — reducing it decreases token consumption per task
  4. Set a monthly budget cap for high-consumption Bots

Security Auditing

  1. Regularly review permission denial records in Security Operations
  2. Analyze denied operations — are they normal permission boundaries or abnormal behavior
  3. Adjust permission configurations or system prompts based on your analysis

Troubleshooting

  1. Filter by “Task Failed” type in Audit Logs
  2. View the details of failed tasks to understand the failure reason
  3. Check tool call parameters and return results
  4. Adjust Bot configuration or resource bindings based on the failure reason

FAQ

Cost statistics don’t match actual API charges

The platform’s cost statistics are estimates based on token counts and model pricing. Actual API charges may differ due to varying billing rules across providers.

Too many logs — how to find key information

Use filtering to narrow down the scope: first select the Bot, then select the operation type, and finally limit the time range. Focus on “Permission Denied” and “Task Failed” log types.

How to restore after budget is exhausted

  • Increase the monthly budget amount in Bot settings
  • Or wait for the automatic reset next month
  • Setting the budget to 0 means no budget limit