Intel / article

You Said No MCP

Source: earendil.com
Author: yarapavan
Date: Published Sep 30, 2026

Context

Pi, an open-source coding and extensible , now includes built-in support after having said it did not support it. MCP connects agents to external tools; in Pi those tools are exposed inside a JavaScript called Codemode, where an agent can use code to sequence calls and combine their results. The developers say the MCP of today differs from earlier versions and that rethinking it is why it is now part of the core. The developers explain the reversal.

The change also required richer configuration so that tools can be marked as deferred, loaded only when needed, or available only through Codemode, which keeps the model from seeing more than it should. Bringing that machinery into the agent's core makes tool discovery and composition part of one design. The developers argue MCP should work more like OpenAPI, with tools that return structured data and can be discovered through their documentation, and they name composability as MCP's long-standing weakness. Design rationale.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.

Read the source earendil.com