Analytics & Monitoring
Monitor workspace performance, track usage, and understand costs with built-in analytics dashboards.
Accessing Analytics
Navigate to workspace → Analytics tab
You'll see an overview dashboard with key metrics and charts.
Key Metrics (KPI Cards)
Total Conversations
- Number of unique conversations started
- Includes all agents in workspace
- Resets based on date filter
Total Messages
- Total messages exchanged (user + agent)
- User messages + AI responses
- Shows engagement level
Total Tokens Used
- Combined input + output tokens
- Varies by model and conversation length
- Used for cost calculation
Total Spend (If credits enabled)
- Total credits spent on AI usage
- Includes platform markup + your markup
- Hidden if
requireCredits = false
Credits Remaining (If credits enabled)
- Current workspace credit balance
- Color-coded: Green (healthy), Yellow (low), Red (empty)
- Hidden if
requireCredits = false
Usage Trends Chart
Line chart showing usage over time:
- Messages - Daily message volume (blue line)
- Tokens - Daily token usage (green line)
- Cost - Daily spending (orange line, if credits enabled)
Use to:
- Spot usage spikes or drops
- Identify busy/quiet periods
- Forecast future usage
Date filters: Last 7 days, 30 days, 90 days, All time
Agent Performance Table
Shows per-agent breakdown:
| Column | Description |
|---|---|
| Agent Name | Name of the agent |
| Conversations | Number of conversations |
| Messages | Total messages sent |
| Avg. Tokens/Msg | Average tokens per message |
| Total Cost | Total spent (if credits enabled) |
Use to:
- Identify most-used agents
- Find inefficient agents (high tokens/msg)
- Allocate resources based on usage
Cost by Agent Chart (If credits enabled)
Bar chart showing spend per agent:
- Each bar = one agent
- Height = total cost
- Sorted by highest spend first
Use to:
- Identify expensive agents
- Optimize model selection for high-cost agents
- Budget planning
Hidden when requireCredits = false.
Date Filtering
All analytics support date range filtering:
Quick filters:
- Last 7 days (default)
- Last 30 days
- Last 90 days
- All time
Custom range:
- Click date picker
- Select start and end date
- Apply filter
All charts and metrics update automatically.
Understanding the Data
What Counts as a Conversation?
A conversation is a unique thread between a user and agent:
- Starts with first user message
- Continues until session ends (30 min inactivity)
- New session = new conversation
How Are Tokens Calculated?
Tokens are units of text:
- ~1 token = 4 characters (English)
- Input tokens: User message + system instructions + conversation history
- Output tokens: Agent's response
Example:
- User: "What are your hours?" (5 tokens)
- System instructions: 100 tokens
- History: 50 tokens
- Agent response: "We're open 9-5 Mon-Fri" (8 tokens)
- Total: 163 tokens (155 input, 8 output)
How Are Costs Calculated?
Cost depends on:
- Tokens used (input + output)
- Model pricing (varies per model)
- Platform markup (20% default)
- Your markup (if workspace credits enabled)
See Credits & Billing for detailed examples.
Monitoring Best Practices
Daily checks:
- Glance at KPI cards for anomalies
- Check Credits Remaining (if enabled)
- Verify no service interruptions
Weekly reviews:
- Review Usage Trends for patterns
- Check Agent Performance for outliers
- Identify optimization opportunities
Monthly analysis:
- Compare month-over-month growth
- Review cost trends
- Adjust pricing/models if needed
Common Insights
High Token/Message Ratio
What it means: Agent is verbose or processing large context
Solutions:
- Reduce conversation history limit
- Shorten system instructions
- Use more concise model (gpt-4o-mini)
Sudden Usage Spike
What it means: New user traffic or viral content
Actions:
- Check credit balance
- Review agent is handling load well
- Consider scaling up if sustained
Low Conversation Count
What it means: Low adoption or poor discoverability
Solutions:
- Promote agent more
- Improve agent placement/visibility
- Review agent's usefulness for users
Cost Growing Faster Than Usage
What it means: Users asking longer questions or agent using expensive model
Actions:
- Review recent conversations
- Consider cheaper model
- Optimize system instructions
Exporting Data
Currently, analytics are view-only in dashboard.
For exports:
- Use API (coming soon)
- Contact support for CSV export
- Take screenshots for reports
Privacy & Data Retention
What's stored:
- Message content (user + agent)
- Conversation metadata (timestamps, agent, model)
- Usage metrics (tokens, costs)
How long:
- Conversations: 90 days (configurable)
- Analytics summaries: Forever
- Transaction history: Forever
Who can see:
- Admin - All analytics including costs
- Editor - All analytics including costs (if no credits)
- Viewer - All analytics except costs
Troubleshooting
Analytics not updating?
- Refresh page
- Check date filter isn't too narrow
- Wait 5 minutes (some metrics batched)
Missing data?
- Verify agent has had usage
- Check correct workspace selected
- Ensure date range includes activity
Costs seem wrong?
- Review transaction history for details
- Verify model pricing hasn't changed
- Check markup percentages in settings
Next Steps
Congratulations! You've completed the Getting Started guide.
Next recommended reads:
- Deployment - Share agents with end users
- Best Practices - Optimize performance and costs
- API Reference - Integrate programmatically