PointGuard AI Launches Gateway for Enterprise AI Workloads with Request Inspection and Cost Control
PointGuard AI Gateway combines access to approved models, prompt inspection, and budgets in a single service. We break down what this provides and what remains behind the scenes.
On October 30, 2026, PointGuard AI announced the launch of PointGuard AI Gateway. The product combines three things in a single service: access to approved models, inspection of requests for sensitive data, and cost control. According to the company, the gateway is also linked to its existing tool-security gateway (a service that checks which external tools an AI agent accesses).
An AI gateway is an intermediate layer between an application and an AI service. It applies rules to requests: where to route them, what to strip out, how much to allow to be spent. Previously, such functions were often assembled piecemeal: routing in one place, budgets in another, data inspection in a third. PointGuard offers a single point where these rules are applied both to working with models and to agent actions. Similar capabilities already exist in LiteLLM: routing, access management, and usage accounting. That is, buyers have something to compare with, and PointGuard's offering depends on how well its security platform ties these functions together.
According to the company's description, the gateway includes routing across approved models, automatic failover (switching to a backup service if the primary is unavailable), prompt inspection, sensitive data protection, caching, and budgets. In practice, this means a team can define rules once and apply them to all applications that go through the gateway. But requesting a response and performing an action require different permissions. An application may have the right to summarize customer data but not the right to send emails or change records. Therefore, it is important whether the permissions, reliability requirements, and responsibility of each scenario are preserved when requests go through a shared layer. This is precisely what the company calls the main question. There is no public data on how the gateway behaves under load or what its limitations are.
It is unknown how much PointGuard AI Gateway costs and what the licensing terms are. There is also no data on exactly how the gateway separates permissions between different applications and agents under shared rules.