AI Agents, MCP and File Transfer

AI agents increasingly produce files (builds, reports, datasets) and need to receive them, but they often run in sandboxes, remote containers or chat interfaces where “download this file” is not an option. The Model Context Protocol (MCP) is the common way to give an agent tools for that, and agent-to-agent protocols let agents exchange messages and files with each other.

The pages build on each other. The explainers are vendor-neutral and apply to any MCP server: what MCP is, local vs remote servers and MCP security. The SRIFT pages then show one implementation: the SRIFT MCP file transfer server, its setup in Claude, Cursor and other clients, and how coding agents use it to deliver files. The last step is agents working with each other rather than with a person: agent-to-agent file transfer explains the problem in general, and AgentNet is SRIFT's answer to it.