AI Agent Security: The Runtime Surface Around Every Tool Call

Originally published at AI Agent Security: The Runtime Surface Around Every Tool Call on smartgate.network.

A shorter version of "AI Agent Security: The Runtime Surface Around Every Tool Call" — the full piece lives at smartgate.network.

What the full piece covers

• ai agent security: the runtime surface, not the input filter — An agent is not a chatbot with more features.
• mcp security: four attack classes the protocol names itself — You do not have to invent the MCP threat model: the protocol's own Security Best Practices document names the attack classes, and every one is a runtime problem.
• mcp server security: what a server has to refuse — The client side asks "can I trust this server".
• prompt injection protection: why the boundary cannot be the prompt — Prompt injection is the vulnerability class where the caller supplies text the model then treats as instruction.
• mcp security best practices, in the order they pay off — The controls above are cheaper in one sequence than in another: each step makes the next one cheaper, because it produces the artifact the next step reads.
• model context protocol security: transport, session, and authorization — The protocol-level surface has three layers, and each one has a specific default you should verify rather than assume.
• mcp gateway security: one enforcement point, one log — If the permission boundary, the server allowlist and the credential scope are enforced in three different places, you have three places to be wrong and no single answer to "what can this agent do".
• mcp prompts: the user-controlled primitive — Tools are chosen by the model; prompts are chosen by the person.
• Where SmartGate fits — SmartGate is the enforcement point this page describes: the place where a tool call is attributed to an identity, checked against a scope, charged against a limit, and written into a record — the same object that the per-key rate …
• Limitations — This page is a control map, not a certification and not a benchmark.

Read the full piece: AI Agent Security: The Runtime Surface Around Every Tool Call on smartgate.network.