F5 AI Gateway tackles the hidden cost problem

Enterprise AI is moving into a new phase. The challenge is no longer just getting models into production. It is keeping track of what those models, agents and tools are doing — and what they are costing.

F5 has enhanced its F5 AI Gateway and integrated it into the F5 AI Security Platform. The aim is to give enterprises a central control layer for AI requests across models, clouds, agents and applications.

The move comes as AI inference becomes a bigger part of enterprise activity. F5’s 2026 State of Application Strategy Report, cited in the announcement, says 77% of organisations now see inference as their dominant AI activity, while organisations manage an average of seven AI models.

AI traffic needs more than monitoring

F5 argues that many enterprises still depend on separate proxies and monitoring tools that were not designed specifically for AI traffic. That creates gaps in governance, security and cost visibility.

The F5 AI Gateway brings three functions together: a Model Gateway for model access and cost optimisation, an MCP Gateway for agent-to-tool governance, and AI Guardrails for protecting prompts and responses.

Budgets, routing policies and agent access controls can be set centrally and enforced across distributed environments. This creates a common operating layer for AI platform, security and finance teams.

The cost of every AI request

Token usage is becoming a key part of AI economics. F5 says its Model Gateway can track tokens by provider, model, team and user, while team budgets can enforce limits as spending happens.

The gateway also supports smart routing, model tiering, semantic caching and GPU-aware load balancing. F5 says these capabilities are designed to reduce token spend by up to 60% without application changes.

Bringing agents under tighter control

The other challenge is agent access. As AI agents interact with APIs, data sources and RAG systems, enterprises need to know what each agent can access.

The MCP Gateway provides fine-grained access controls, an MCP server registry and audit trails showing what was accessed, when, by which agent and on whose behalf.

AI Guardrails add another layer by inspecting prompts and responses, redacting sensitive information and blocking injection and jailbreak attempts.

Read More: 

Xebia and OpenAI partnership takes enterprise AI further

Beyond implementation: How AI could reshape Liferay partner revenue

Rapoo appoints Trustech AV Solutions to push enterprise collaboration

Siliconops.ai launches siliconcloud.in as India eyes sovereignty