Usage Tracking

Monitor API calls, view rate limits, and analyze tool usage

Usage Tracking

Metrifyr tracks every API call made through your MCP server, giving you visibility into usage patterns, rate limits, and tool performance.

What Gets Tracked

Every MCP request logs:

• Timestamp - When the request was made
• Tool name - Which tool was called (e.g., ga4_run_report)
• API key - Which key made the request
• Success/failure - Whether the request succeeded
• Response time - How long the request took
• Error details - If failed, what went wrong

Dashboard Usage Tab

Access usage analytics at Dashboard > Usage.

Overview Metrics

Today:

• Total requests
• Successful requests
• Failed requests (with error count)
• Rate limit consumption (X / 1,000)

This Week:

• Daily request chart
• Most used tools (top 5)
• Error rate trends
• Peak usage times

This Month:

• Total requests
• Average requests per day
• Top tools by request count

Request Chart

Visual timeline showing:

• Daily request volume
• Success vs error distribution
• Trend line (increasing/decreasing usage)

Hover over any day to see detailed breakdown.

Tool Analytics

See which tools you use most:

Click on any tool to see:

• Request history for that tool
• Common errors
• Response time distribution

Filter by API Key

If you have multiple API keys, filter usage by key:

1. Select a key from the dropdown
2. Usage stats update to show only that key
3. See which tools each key uses most

Useful for:

• Tracking different projects separately
• Identifying which keys hit rate limits
• Auditing team member usage

Rate Limits

Metrifyr enforces rate limits to prevent abuse and ensure fair use.

Current Limits

Free tier:

• 1,000 requests per day
• Resets at midnight UTC
• Shared across all API keys on your account

Pro tier (coming soon):

• 10,000 requests per day
• Per-key limits (each key gets its own quota)
• Priority support

Checking Your Limit

In dashboard:

• Usage tab shows "X / 1,000 requests used today"
• Progress bar indicates consumption
• "Resets in X hours" countdown

In MCP responses:

• Rate limit headers included in every response
• X-RateLimit-Remaining: 847
• X-RateLimit-Reset: 1642348800

What Happens When You Hit the Limit?

Soft limit (90%):

• Warning banner appears in dashboard
• Email notification sent
• Requests still work normally

Hard limit (100%):

• New requests return 429 Too Many Requests
• Existing requests in flight complete normally
• Limit resets at midnight UTC

Upgrade prompt: Contact support to upgrade to Pro tier for higher limits.

Optimizing Usage

Tips to reduce requests:

1. Batch queries - Ask for multiple metrics in one query instead of separate requests
2. Use caching - Claude caches recent responses, re-use them when possible
3. Set defaults - Use Memory to avoid repeated property lookups
4. Filter by key - Disable unused tools to prevent accidental usage

See best practices guide

Error Tracking

The usage tab shows error statistics:

Error Rate

Overall:

• Success: 1,847 (98.2%)
• Errors: 34 (1.8%)

By tool:

• ga4_run_report - 2% error rate
• gsc_query_search_analytics - 0% error rate

Common Errors

Top errors with counts:

1. 401 Unauthorized (12) - Invalid API key or revoked OAuth
2. 403 Forbidden (8) - Missing OAuth scope
3. 404 Not Found (7) - Property or account doesn't exist
4. 429 Too Many Requests (4) - Rate limit hit
5. 500 Internal Server Error (3) - Server issue

Click on any error to see:

• Full error message
• When it occurred
• Which tool/key triggered it
• Suggested fix

Error Resolution

401 Unauthorized:

• Check API key is correct
• Verify Google account is still connected
• Reconnect if needed

403 Forbidden:

• Grant missing OAuth scope
• Reconnect Google account with broader permissions

404 Not Found:

• Verify property/account ID is correct
• Check you have access to the resource

See troubleshooting guide

Response Time Analytics

Track how fast tools respond:

Average Response Times

Percentiles explained:

• P50 (median) - 50% of requests are faster than this
• P95 - 95% of requests are faster than this
• P99 - 99% of requests are faster than this

Slow Query Detection

If a tool is consistently slow (>5s avg):

• Warning appears in dashboard
• Email notification sent (optional)
• Suggested optimizations shown

Common causes:

• Large date ranges
• Complex report queries
• Google API throttling

Optimizations:

• Reduce date range (e.g., 7 days instead of 90)
• Request fewer metrics/dimensions
• Use pagination for large result sets

Export Usage Data

Export your usage data for external analysis:

1. Go to Usage tab
2. Click "Export"
3. Choose format (CSV, JSON)
4. Select date range
5. Download file

Exported data includes:

• Timestamp
• Tool name
• API key ID (hashed)
• Success/failure
• Response time
• Error details (if failed)

Use cases:

• Import into spreadsheet for analysis
• Generate custom reports
• Audit tool usage for billing
• Debug patterns in your integration

Usage Alerts (Coming Soon)

Set up alerts for usage events:

Alert types:

• Rate limit warnings (80%, 90%, 95%)
• Error rate spikes (>5% errors in 1 hour)
• Slow query detection (>10s avg for 30min)
• Daily usage summaries

Delivery methods:

• Email
• Slack
• Webhook (POST to your endpoint)

Next Steps