Cisco looses Splunk to probe its growing agentic menagerie

"The job of agentic babysitter goes to a new AI Agent Monitoring tool for Splunk's Observability Cloud. We're told it visualizes agent workflows and can track the performance, cost, quality, and behavior of LLM and agentic applications. Splunkers can take it for a spin two weeks from now. Cisco will integrate the new Splunk tool with AI Defense, the suite it announced last year, which makes sure LLMs don't misbehave or create risks."

"The networking giant made AI Defense generally available this week and celebrated that milestone by adding extra capabilities, including the ability to catalog of Model Context Protocol (MCP) servers an enterprise uses in-house and elsewhere. Again the plan here is helping users to spot systems that might have escaped their policies and AI guardrails. Cisco also added automated red-teaming tools to test the security of models and associated infrastructure."

"Switchzilla designed this one to run on prem so queries and prompts don't have to leave the building, but it can still target off-prem resources. And in a nod to the necessity of complying with older operational models, Cisco has mapped AI Defense to AI frameworks from NIST, OWASP, and MITRE. The company has also announced plenty more agents to automate campus, branch, and industrial networks."

Cisco plans to deliver Cloud Control later in 2026 while releasing multiple agentic and governance tools beforehand. A new AI Agent Monitoring tool for Splunk Observability visualizes agent workflows and tracks performance, cost, quality, and behavior of LLM and agentic applications, with availability in two weeks. Cisco integrated the Splunk tool into AI Defense, which is now generally available and adds capabilities such as cataloging MCP servers and automated on-prem red-teaming to test model and infrastructure security. AI Defense has been mapped to NIST, OWASP, and MITRE frameworks. Additional agents will automate campus, branch, and industrial network troubleshooting and continuous optimization by mid-year.

Read at Theregister