Introduction to Splunk AI Agent Monitoring
Monitor and troubleshoot the performance, quality, token usage, estimated cost, and risk of your AI agents and applications.
Monitor and troubleshoot the performance, quality, token usage, estimated cost, and risk of your AI agents and applications with Splunk AI Agent Monitoring.
Cost is estimated by multiplying the published provider cost by the number of available tokens. The cost estimation doesn't reflect the actual billing cost of your AI agents and applications.
Get started with AI Agent Monitoring
You can ingest data from AI agents and applications using the following methods:
-
Instrument your AI application with zero-code instrumentation for supported frameworks and code-based instrumentation for other AI applications.
-
Translate and collect data from AI applications that you've already instrumented with a supported third-party instrumentation library.
To get started, see Set up AI Agent Monitoring.
What can I do with Splunk AI Agent Monitoring?
After you set up data collection from AI agents and applications, the data populates built-in experiences that you can use to monitor and troubleshoot them.
| Do this | With this tool |
|---|---|
| Monitor your overall AI application and agent environment. | AI overview page |
| Monitor your AI agents. | AI agents page |
| Monitor the traces and spans associated with your AI agents. | AI trace data page |
| View all of your LLM service dependency graphs and user interactions. | Service map |
Supported AI instrumentation libraries
Splunk Observability Cloud supports all instrumentation libraries available with the OpenTelemetry GenAI utility. For more information about the supported instrumentation libraries, see the opentelemetry-util-genai repository on GitHub.
Splunk Observability Cloud follows the semantic conventions for supported instrumentation libraries. For more information, see Semantic conventions for generative AI systems in the semantic-conventions-genai GitHub repository.