What is an MCP Tool Call?
A JSON-RPC request from an AI agent to execute a specific function exposed by an MCP server, containing the tool name, arguments, and optional metadata.
WHY IT MATTERS
Every action an AI agent takes through MCP is a tool call. Reading a file, sending an email, creating a database record, executing a payment — each is a discrete tool call that the server processes and returns a result for.
Tool calls are the atomic unit of agent behaviour. They're also the atomic unit of risk. A single malformed or malicious tool call can exfiltrate data, delete records, or trigger financial transactions. This makes the tool call the natural enforcement point — if you can inspect and gate every tool call, you can control what an agent does.
HOW POLICYLAYER USES THIS
Intercept inspects every MCP tool call in transit, evaluating it against YAML-defined policies before forwarding to the upstream server. Rate limits, category restrictions, and argument validation all operate at the tool call level.