Discussion on "What Is MCP (Model Context Protocol) and Why It Matters for Agents"

Nagappan S · 2026-09-29T05:31:53.582Z

You've probably built at least one integration where your AI app needed to talk to Slack, then Google Drive, then your internal database, then GitHub — and each one meant writing a different client, h

The split you draw between resources being safe to read and tools needing care is the part most MCP intros skip. In practice that boundary is only as good as the server's honesty about which calls mutate state, and I have seen a "read" tool quietly write. Do you gate tool invocation behind an approval step, or trust the server's own tool-vs-resource labeling?

That "read tool that quietly writes" case is exactly why I wouldn't trust the labels. Annotations like a read-only hint are self-reported by the server, so I treat them as UI hints, not a security boundary.

My approach: require human approval by default for anything that can mutate state or reach external systems, auto-approve only an allowlist of tools I've reviewed, and enforce the real boundary below the protocol, with read-only credentials for read use cases, so a mislabeled tool can't write even if it tries. Resources vs tools is a useful design signal, but permissions have to be enforced where the data lives.

Have you found anything that catches mislabeled tools before they run, or is it mostly review and logging?

Good overview of the MCP building blocks. One practical angle worth exploring: how MCP handles the payment step in agent workflows. When an agent discovers a tool through MCP that charges for usage (API credits, compute, data access), the protocol currently has no native way to pass payment context between the client and server. A few teams are experimenting with extending the tool primitive to carry payment parameters, which would let agents settle costs autonomously without leaving the MCP interface.

Search Hashnode

What Is MCP (Model Context Protocol) and Why It Matters for Agents

Responses(3)