v1.3.0
OpenCodeSummary
This release introduces major new capabilities, including full support for the GitLab Agent Platform and git-backed session review modes. Users will also benefit from enhanced security with support for multistep authentication flows directly in the desktop app. Several critical bug fixes address issues with remote server switching, Windows command execution, and terminal stability.
New Features
- Enabled full support for the GitLab Agent Platform, including automatic discovery of workflow models from your GitLab instance.
- Introduced git-backed session review modes, allowing users to review uncommitted changes and branch diffs directly within opencode.
- Added support for multistep authentication flows in the desktop app, enabling login via OAuth providers and SAML.
- Implemented an interactive update flow that shows a confirmation dialog before major/minor upgrades, allowing users to review changes or skip versions.
- Enabled opencode to run on Node.js in addition to Bun, with a dedicated entry point and build script.
Bug Fixes
- Fixed stale server URL and email appearing on re-login after switching to a different server environment.
- Fixed issues when switching between remote servers that caused stale project data, broken terminal spawning, and incorrect sidebar project display.
- Fixed Windows command execution for shims (.cmd/.bat files) which previously failed during upgrades, uninstalls, IDE installs, and LSP launches.
- Fixed tool discovery for npm-installed packages by correctly including the cache bin directory in PATH lookups.
- Fixed tool and session error messages to display in a clean, human-readable format instead of confusing technical strings.
- Fixed issue where terminals became unresponsive when the server returned a "not found" error for terminal sessions; terminals now properly reinitialize.
- Fixed paste operations to preserve raw byte data, ensuring proper handling of binary content and special characters when pasting into the terminal.
- Fixed issue where rapidly pressing Enter could submit prompts multiple times; the input now ignores repeated Enter submits.
- Fixed Vertex AI errors with non-thinking models (like gemini-2.0-flash) by correctly applying thinking configuration only to reasoning-capable models.
- Fixed streaming error where cost data was sent in upstream provider format instead of client-requested format, preventing validation errors.
- Fixed review file comments to submit when clicked and display without horizontal overflow, matching diff review behavior.
- Fixed file and image attachments being silently lost when reusing or forking a previous prompt that contained attachments.
- Fixed snapshot revert issues where overlapping file changes could apply the wrong patch hash and resolved a sorting bug in hidden file ordering.
- Fixed terminal renaming from the context menu, which previously failed due to dropdown focus conflicts.
- Fixed console device login to properly handle pending authorization states, ensuring the login process continues polling instead of failing while waiting for approval.
Improvements
- Enabled faster project navigation using keyboard shortcuts (cmd+option+arrow keys) in the desktop app.
- Desktop app now supports file type filters and multiple file selection when uploading files.
- Users can now access the review panel before starting a session, showing proper empty states when no changes exist.
- Fixed visual flickering when switching between workspaces or projects; the interface now transitions smoothly without showing loading spinners.
- Fixed issue where the sidebar would automatically close when resized to minimum width; it now clamps at minimum width instead of closing.
- Improved loading state management across the app for smoother, more consistent feedback during asynchronous operations.
- Improved xAI reasoning models performance in long multi-turn conversations by switching to the responses API with encrypted content support.
- Updated the GPT subscription model list with the latest available models.
- Improved error detection for vLLM context overflow errors, providing clearer feedback when context limits are exceeded.
- OAuth plan users can now access codex instructions freely without previous whitelisting restrictions.
- Improved UI contrast and reduced brightness of hover/active elements for better readability and reduced eye strain.
- Implemented a streamlined fallback system prompt that removes unnecessary boilerplate text for cleaner AI responses.
- Stabilized the order of agents and skills in system prompts via improved prompt caching for more consistent AI responses.