GitHub Copilot July 2026: Codex Lands in JetBrains

GitHub keeps shipping Copilot updates weekly, and the July 7 changelog is one of the bigger ones: Codex arrives as an agent provider inside JetBrains IDEs, Claude sessions get real permission controls, and Inline Chat finally goes GA. If you're running IntelliJ, PyCharm, or WebStorm with Copilot installed, there's a config change worth making today.

Codex Comes to JetBrains (Public Preview)

You can now run OpenAI's Codex CLI as an agent provider without leaving your JetBrains IDE. Install the Codex CLI, point Settings > Tools > GitHub Copilot > Chat at the CLI path, then pick Codex from the agent picker in the Copilot Chat panel. It's public preview, not GA — Business and Enterprise admins need to flip the editor preview features policy before their teams get access.

Claude Sessions Get Permission Modes and Debug Logs

Claude agent sessions in Copilot now support permission mode selection from the chat input dropdown, so you choose how much autonomy the agent gets per session instead of accepting a fixed default. Claude sessions also now show up in agent debug logs, which matters if you've been debugging why an agent run stalled or made an unexpected change with no visibility into what happened.

Autopilot, Hooks, and MCP Management Level Up

Copilot CLI sessions get new approval tiers: Default, Bypass Approvals (all tool calls auto-approved, no dialogs), and Autopilot Preview (auto-approve plus auto-response to clarifying questions, so the agent keeps iterating instead of stalling for input). The Agent Customizations editor now handles Hooks directly, and MCP servers can be browsed, installed, started, stopped, and uninstalled from the same view — including workspace-level servers defined in .github/mcp.json. Inline Chat, meanwhile, is generally available after its private beta.

Why This Matters in the Three-Horse Race

Copilot's professional-developer market share has slipped from 67% to 51% over the past year as Cursor and Claude Code eat into its lead. Multi-agent support — Codex, Claude, and Copilot's own models all selectable from one picker — is GitHub's answer: don't force a single model choice, let developers route each task to whichever agent handles it best inside the IDE they already use. That's a different bet than Anthropic's, which is doubling down on Claude Code as a standalone terminal-native agent following its own SpaceX-driven Claude Code rate limit increase. It's worth checking your admin settings against Copilot's usage-based AI Credits billing before turning on Codex or extra agent sessions — more agent options means more surfaces quietly consuming your credit pool.

Sources: GitHub Changelog, TechTimes.