Kong announced the general availability of AI cost-management features on November 4, 2026, for businesses that need to connect AI spending to the teams, applications and agents responsible for it. The company says the tools in Kong AI Gateway support cost allocation and budgeting across AI models and providers.
The company also announced expanded agent-session tracing in its Konnect management platform. A session is the connected set of steps an AI application takes to complete a task. Kong says the session view connects model calls, tool use and policy checks so teams can investigate cost, delays and errors.
Access to that session view needs checking. Kong had already previewed it in September, and that earlier product post still describes it as coming soon. The November announcement does not establish that every customer can enable it today.
Who incurred the cost, and where did it build up?
For business owners, the two capabilities answer different questions. A team breakdown can support allocating expenses to departments; a session trace can help locate costly or slow steps within a task. Neither, by itself, establishes whether the result was useful or worth the money.
A gateway sits between an application and its AI services. Kong’s cost-reporting interface covers gateway usage and separately submitted usage. Its documentation says activity without an accepted identity remains unattributed. The interface is still labelled beta, despite the broader product’s general-availability announcement.
The practical implication is to check which applications feed the report and how their usage maps to the right team. Keep unassigned spending visible; do not assume the report captures every company AI subscription or external tool bill.
A reporting budget does not itself stop requests
Kong’s API specification makes a consequential distinction: a budget’s posture field records whether spending should continue or stop at the limit, but does not itself block requests—even when set to enforce.
Kong documents a separate cost-based rate-limiting policy for controlling requests. It uses the gateway’s calculated cost, which depends on configured model prices, including relevant cached-token and service-tier rates. Teams should check those settings and reconcile reported amounts with their billing terms before allocating charges internally.
There is another reporting detail worth checking: the usage-event specification treats an omitted cost as unknown but displays it as zero in event lists. A displayed zero therefore need not mean the usage was free.
Check the features included in the price
Kong’s AI Management pricing page advertises Plus from US$25 a month plus usage. It explicitly lists AI cost management under Enterprise, which has custom annual billing. The starting price is not confirmation that the announced cost-attribution and session-tracing features are all included.
Before buying, confirm access to the required features, the usage that will be captured and which controls actually stop spending. Evaluate the platform’s own charges and operating effort alongside any prospective savings; the launch does not demonstrate a reduction in a customer’s total costs.
For businesses chiefly trying to allocate Claude spending, RohitAI’s guide to Claude’s own cost analytics examines a narrower reporting option. The reason to consider Kong is a missing cross-provider allocation or investigation capability—not simply the availability of another dashboard.