GitHub Copilot for JetBrains 1.18.0 Adds AI-Assisted Tool Approvals and Editable Agent History
GitHub Copilot for JetBrains 1.18.0 shipped a batch of features focused on approval control, conversation editing, and organizational consistency.
Details
- AI-assisted tool approvals (preview): automatically approves low-risk tool calls while still requiring confirmation for higher-risk actions, cutting approval interruptions without giving up control over sensitive operations
- Editable agent history: users can re-edit previous messages in an agent session, rewinding both conversation history and file modifications to revise an earlier request without cluttering the chat
- Org-level customization: support for organization and enterprise-level skills and custom instructions across both local and agent sessions, for consistent team practices
- Codex agent plan mode: lets developers review and approve a suggested approach before the agent implements changes
- MCP tool management: granular, persistent per-tool settings for Model Context Protocol tools, plus a toggle for the built-in GitHub MCP Server
- Interface changes: a side-by-side chat panel switcher for browsing and conversing simultaneously, plus improved discoverability of commands, customizations, and settings, alongside reliability fixes for inline chat and Codex sessions
What happened next
The release leans toward giving developers more control over agent behavior — approving what gets auto-run, editing history instead of restarting a session, previewing a plan before execution — rather than adding new raw capability, consistent with GitHub’s recent focus on trust and governance features across Copilot surfaces.