Stateless MCP Protocol Revives Interest in Tool Integration for LLMs
The Model Context Protocol (MCP) has seen a resurgence of interest following the release of its 2.0 specification on July 28, 2026. This stateless version significantly simplifies how applications integrate with large language model agent frameworks.
Background on MCP
First introduced by Anthropic in November 2024 as a way to standardize tool exposure for LLM agents, MCP initially gained traction before being somewhat eclipsed by Anthropic’s Skills system. However, the new stateless design addresses many of the earlier limitations and offers compelling advantages.
The primary benefit is enhanced security—giving agents access to tools through audited APIs like MCP is safer than providing unrestricted shell environments with internet access.
Key Improvements in Stateless MCP
With the updated protocol, all interactions occur within a single HTTP request:
POST /mcp HTTP/1.1
Content-Type: application/json
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "search",
"arguments": {
"q": "otters"
}
}
}
This contrasts with the previous stateful approach that required two requests—one to initialize a session and another to call the tool.
The stateless design eliminates server-side session management, making it easier to build scalable applications while ensuring better auditability and control over agent interactions.
New Tools for MCP Exploration
To facilitate experimentation with the protocol, I’ve developed two new open-source tools:
- mcp-explorer: A CLI tool that allows users to interactively discover and test available MCP endpoints (https://github.com/simonw/mcp-explorer)
- datasette-mcp: A Datasette plugin for serving MCP tooling through a familiar interface (https://github.com/simonw/datasette-mcp)
The mcp-explorer can be used with uvx to list tools from any public MCP server:
uvx mcp-explorer list https://agentic-mermaid.dev/mcp```
This command will return a comprehensive listing of available tools on Ade Oshineye's demo server, showcasing the protocol in action.
The revival of MCP with its stateless design presents a more secure and manageable approach to integrating specialized functionality into LLM agent workflows.