Linear integration, improved Agent terminal, and OS notifications

Source: Cursor Date: 2025-08-21 URL: https://cursor.com/changelog/1-5

Summary

Cursor 1.5 ships Linear integration so Background Agents can be launched directly from Linear issues, native OS notifications when agent runs complete or require input, and MCP elicitation support enabling servers to request structured user input via JSON schemas. Agent terminal gets left-side positioning with animation improvements, and Jupyter notebooks gain full-context Tab awareness. GPT-5 with Agent now supports todo creation.

Implications

Agent-IDE feature race. Linear integration is a significant workflow bridge — agents now attach to issue trackers, not just files. This mirrors what GitHub Copilot Workspace does with GitHub Issues and what Claude Code does via gh CLI. When the agent loop originates in a task system rather than a file, the IDE becomes an execution layer for project management, not just an editor. Aider and Codex CLI lack this native integration.

MCP/protocol adoption. MCP elicitation support (structured user input via JSON schema) is Cursor signaling it’s treating MCP as a first-class protocol, not a bolt-on. This matters for the downstream tool ecosystem: MCP server authors can now build richer interactive agents inside Cursor without custom UI.

Model integration cadence. GPT-5 gaining todo creation within the agent loop shows Cursor is exposing model-specific capabilities rather than abstracting them away — model choice has UX consequences, which locks users into preferred providers.