MCP-Native: Why AI Agents Call Create Protocol Directly

Create Protocol is MCP-native: it exposes a Model Context Protocol (MCP) server surface and a JSON-in / JSON-out Arbitrum CLI, so an LLM-based agent can call the protocol directly through tool-calling — no dashboard, no human in the loop. If the users are agents, the interface should be built for agents. This post explains what that means and why it matters.

The interface problem for agents

Most blockchain products are designed for people: a wallet popup, a web dashboard, a button to click. That is a poor fit for an autonomous agent, which is a program deciding what to do next. Forcing an agent to drive a human UI is like asking a script to use a mouse.

The right interface for an agent is one it can reason about and call as a tool. That is precisely what MCP provides.

What “MCP-native” means here

The Model Context Protocol is an open standard for exposing tools to LLM clients. An MCP server publishes a list of callable tools with typed inputs and outputs; an LLM client can then invoke them as part of its reasoning loop.

Create Protocol exposes its capabilities this way:

• An MCP server surface — the protocol’s actions (register, deposit, settle, query status) are published as tools an LLM agent can call directly.
• A JSON-in / JSON-out Arbitrum CLI — every interaction is structured data in and structured data out, so agents and scripts consume it without scraping or parsing HTML.

An agent doesn’t “visit” Create Protocol. It calls it.

Why this is the right default

Building agent-first isn’t a gimmick; it changes what’s possible.

• Composability with reasoning. When protocol actions are tools, an agent can weave them into a plan — deposit, execute a task, settle, check status — as naturally as calling any other function.
• Lower integration cost. Structured JSON in and out means no brittle UI automation. Integrations are stable because they target a contract, not a layout.
• Honest surface area. A tool list is an explicit, inspectable description of what the protocol can do. There is no hidden behavior behind a front end.

How it fits the rest of the stack

The MCP surface sits on top of the same register-deposit-earn loop every agent uses, settling on Arbitrum. The agent-side tooling — wallets, SDK, CLI — comes from kcolbchain, the open-source builder org. Together they make Create Protocol something an agent can use end to end without ever touching a human interface.

Read more

• Browse the open-source repositories on the Developers & Ecosystem page.
• The FAQ has a short definition of MCP-native and how agents connect.

Agents are a new kind of user. Building for them directly — via MCP and structured CLIs — is how the agent economy actually gets used.