Skip to main content
You can export HTTP logs from your AI Gateway endpoints to external systems using ngrok’s Log Exporting feature.

​
Current support

AI Gateway endpoints emit standard ngrok traffic logs (http_request_complete.v0). AI-specific log fields (provider selection, retry tracking, token usage) are not currently available but are on our roadmap.
Log Exporting allows you to:
  • Stream request/response logs to your analytics platform
  • Archive traffic data for compliance
  • Build custom dashboards and alerts
  • Integrate with your existing observability stack

​
Setting up log exporting

Log Exports are configured at the account level through the ngrok Dashboard:
1

Open Log Exporting

Go to Log Exporting in the ngrok Dashboard
2

Create a New Log Export

Click to create a new Log Export
3

Select Log Sources

Choose which log sources to capture (for example, http_request_complete.v0 for HTTP traffic)
4

Configure Destinations

Select and configure one or more destinations for your logs

​
Supported destinations

ngrok supports exporting logs to:
  • Amazon CloudWatch - AWS logging and monitoring
  • Amazon Firehose - AWS data streaming (can deliver to S3)
  • Amazon Kinesis - AWS data streaming
  • Azure Logs Ingestion - Azure logging
  • Datadog Logs - Monitoring and analytics
For detailed configuration of each destination type, see the Log Exporting documentation.

​
What logs include

Standard HTTP traffic logs from AI Gateway endpoints include:
FieldDescription
event_idUnique identifier for the log
event_timestampWhen the request completed
account_idYour ngrok account ID
conn.client_ipClient IP address
conn.server_nameEndpoint hostname
http.request.methodHTTP method (POST, GET, etc.)
http.request.url.pathRequest path (for example, /v1/chat/completions)
http.response.status_codeResponse status code
http.response.body_lengthResponse body size

​
What’s not yet included

AI Gateway-specific data is not currently included in logs:
  • Which provider handled the request
  • Model selection strategy results
  • Retry attempts and failover decisions
  • Token counts (input/output/total)
  • Cost estimates
  • Provider API key used
These AI-specific log fields are on our roadmap. Email [email protected] to tell us what data you need - your feedback shapes our development priorities.

​
Filtering logs

You can filter traffic logs using CEL expressions to capture only relevant requests: 
# Only log requests to chat completions endpoint
ev.http.request.url.path == "/v1/chat/completions"

# Only log errors
ev.http.response.status_code >= 400

# Only log slow requests (if timing fields available)
ev.http.response.status_code == 200
See the Log Exporting documentation for more filter examples.

​
Use cases

​
Compliance logging

Stream all requests to long-term storage for audit trails:
  • Use Amazon Firehose to deliver to S3
  • All HTTP requests are captured with timestamps and metadata
  • Combine with Traffic Inspector for full body capture when needed

​
Real-time monitoring

Send logs to your monitoring platform:
  • Use Datadog or CloudWatch destinations
  • Build dashboards showing request volume and latency
  • Set up alerts for error spikes

​
Custom analytics

Process logs in your own systems:
  • Stream to Kinesis or Firehose
  • Build custom pipelines for your specific needs
  • Aggregate and analyze traffic patterns

​
See also