2025-09-12 · Cursor

Slash commands, summarization, and improved Agent terminal

protocolsagents

read at source ↗ cursor.com

Slash commands, summarization, and improved Agent terminal

Source: Cursor Date: 2025-09-12 URL: https://cursor.com/changelog/1-6

Summary

Cursor 1.6 ships custom slash commands (reusable team prompts in .cursor/commands/[command].md, triggered via /), on-demand /summarize for context compression without restarting, and MCP Resources support for server-shared context (files, database schemas). Agent terminal gets major reliability fixes across 45 patches addressing hanging commands, SSH performance, Bash/ZSH state restoration, and WSL compatibility.

Implications

Agent-IDE feature race. .cursor/commands/ as a team-shared prompt library is a direct parallel to Claude Code’s slash commands and Aider’s --read patterns. Storing commands as markdown files in the repo makes them versionable and reviewable — a stronger collaboration model than IDE settings. /summarize for context management shows Cursor is treating context window pressure as a first-class UX problem rather than a model limitation to hide.

MCP/protocol adoption. MCP Resources support (server-shared context like files and DB schemas) extends Cursor’s MCP integration beyond tool calls into data sharing. This closes a gap with how production MCP servers are actually used — many real use cases involve injecting structured context, not just calling tools.

Watch: Whether .cursor/commands/ evolves into a package-like distribution model (npm install cursor-commands) and whether terminal reliability improvements translate to fewer agent abandonments on long-running tasks.

← all signals