The Pulse
Cisco Adds Token-Spend Tracking to Splunk Observability
Cisco is adding Tokenomics capabilities to Splunk Agent Observability to track AI token spending, coding-agent usage and projected consumption. The feature is available across Splunk Observability Cloud, Cisco Cloud Control and Splunk’s on-
AI.info Team ·
“Customers want to know: Can I trust it to do the job? Can I afford it? And, most importantly, can I secure it?”
Jeetu Patel, President and Chief Product Officer, Cisco
Cisco is adding a Tokenomics capability to Splunk Agent Observability, giving companies a way to track AI token consumption, assign costs to users and agents, and estimate where spending will land before a billing period ends.
The announcement, made at Splunk.conf in Denver on September 15, puts token expenditure alongside agent performance and infrastructure telemetry. Cisco says the feature is designed to help companies see how AI costs accumulate rather than discovering them only after an invoice arrives.
Token spend moves into the observability console
Tokenomics is part of Splunk Agent Observability, which Cisco says now runs in Splunk Observability Cloud and Cisco Cloud Control as well as its on-premises deployment. The software evaluates agent and model behavior, monitors performance across the AI stack and applies runtime guardrails intended to block unsafe actions, including sensitive-data leakage and inaccurate outputs.
The new capability focuses on the financial side of those operations. It tracks and attributes token expenditure across AI agents and employee use of coding tools such as Claude Code, Codex and Cursor. Splunk’s accompanying product post says the system surfaces usage, cost, adoption and productivity information at both team and user level.
That reporting matters because employee-facing coding agents can create costs outside the traditional application budget. A company may know which model provider it uses without having a clear view of which teams, projects or individual workflows are driving consumption. Cisco’s release describes Tokenomics as a way to connect that spending with business outcomes.
Forecasting replaces the end-of-month surprise
Tokenomics also forecasts consumption patterns before the billing period closes. Cisco says the projection uses its Deep Time Series Model to estimate where spending is headed, while Splunk says the consolidated view can help organizations route workloads toward a more cost-effective model.
The announcement does not describe a fixed dollar cap or a hard token quota that automatically stops an agent. The controls disclosed by Cisco center on visibility, attribution, forecasting and runtime safeguards. That distinction makes the feature closer to an operating and budgeting layer for AI use than a conventional prepaid usage limit.
One product update among several Splunk changes
Cisco is packaging the Tokenomics announcement with a wider set of Splunk updates. Observability Studio is intended to help developers instrument applications from the start, while a new Network Intelligence App brings Cisco network topology, device health and events into Splunk so teams can trace an alert to the affected device and surrounding network.
Cisco also announced Essentials and Premier editions for Observability Cloud, with both including Splunk Agent Observability, application performance monitoring, infrastructure monitoring and Observability Logs. The company says the editions are intended to simplify how customers buy and expand observability services.
The cost question follows the agent
AI agents introduce a cost pattern that differs from a conventional software request: a single task can trigger repeated model calls, tool use and extended context. Monitoring the agent without monitoring the tokens can therefore leave the most direct measure of operating expense outside the main dashboard.
Splunk’s new view ties those two streams together. Teams can inspect agent behavior, connect usage to users or groups, and project consumption through the current billing cycle. Cisco’s stated objective is less about making AI cheaper in the abstract than giving organizations enough accounting detail to decide where AI use earns its cost.