Urgent.News

What's breaking now, across thousands of outlets.

AI

Pi coding agent pulls a 180 and adds MCP support

To maintain minimalism, a new harness layer has been folded into a separate component

Pi coding agent pulls a 180 and adds MCP support

The Pi coding agent achieved its 1.0 release on Thursday while undergoing treatment for whiplash. One significant update is the addition of Model Context Protocol (MCP) support, initially dismissed by Pi's creator, Mario Zechner, as unnecessary.

MCP's newfound importance stems from its improvements, as well as the enhancements Pi developers incorporated, making it more convenient to incorporate other features. Earendil, the company that acquired Pi in April 2026, highlighted this, noting that both MCP and Pi require a sandbox environment, such as an interpreter.

Earendil describes Pi as a hardened, minimal, and extensible agent harness that users can customize. Comparing Pi to a Python web framework, it would be akin to Flask over Django. Despite its minimalistic nature, Pi has garnered new features, including Codemode, a sandboxed environment for tool calls with MCP, decision models, and image model support; extension support for virtual models; deferred tool loading; cache warming for Anthropic models; mid-conversation system messages; a default full-screen mode; and TUI theme.

Parallel to Pi, Earendil built Pi Durable, a separate substrate designed for long-running agentic applications. It serves as a distinct component, preserving Pi's minimalist design while connecting the underlying model to tools for conversation management. Pi Durable, explained the company, does not replace Pi but acts as a framework for constructing various agentic applications, including coding agents. It shares code and principles with Pi, such as minimalism and malleability.

Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Every outlet that reported this, and when →

Read the original at theregister.com →

More in AI

Time for AI guardrails is right now

The genie is most definitely out of the bottle. And while there’s little chance of putting it back in, we should be very concerned that the people who released it […]

  • AI guardrails urgently needed as AI develops own decision-making
  • Recent breaches show AI agents hacking government systems
  • U.S. President Trump backs unrestricted AI development

More from Friday 2 October →