Xorin releases
Changelog
A compact record of user-visible additions and improvements. Select a release to expand its details.
2.2.1 Latest
- Deeper: Long steps can now run for minutes without timing out. Deeper is still not recommended for big builds. Learn more
- Stop at a limit: Stopping at the spend or turn limit now applies waiting scripts. Learn more
- Fixes: More reliable script changes, Try again and Continue, Play checks, and project search.
2.2.0
- Unity 6.6: Works with Unity 2022.3 LTS up to Unity 6.6. Macs now require Apple Silicon. Learn more
- Thinking level: Choose how much the model thinks: Faster, Balanced, or Deeper. Learn more
- Clearer errors: Plain error messages with an Error ID, Try again, and Continue at the step limit. Learn more
- Models: New models including Claude Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol and Terra, Gemini 4 Argon, Kimi K3.1 Pro, and GLM 5.3. Claude Opus 5.5 is the new default, and the model list can now update without a new release. Learn more
- Workflows: Xorin can run many reads and edits together as one workflow, shown as a single card in chat. Learn more
- Play checks: Play checks can click, press keys, watch values over time, and check sounds and held items. Problems they find stay open until cleared, and answers say when player input was not tested. Learn more
- Clear results: Answers say what changed, what was not tested, and what was left in the scene, with a checklist of what you asked for. Auto-approve is now Auto-keep. Learn more
- Safer changes: Xorin asks before deleting your work, catches an unsaved scene before sending, and repairs compile errors in files it did not write. Editor scripts stop after 30 seconds. Learn more
- Chat: Welcome cards in an empty chat, live progress and cost, the cost of a stopped request, and 👍/👎 ratings on answers. Learn more
- Updates: Xorin tells you when a new version is available and shows what's new after updating. Learn more
- Large projects: Better search and project summaries that say what they left out; new scripts go beside their feature instead of into imported packages.
- Assets: Clearer generation limits, a "not charged" note when generation fails, seamless looping sounds, and sprite-sheet and particle-material fixes. Learn more
2.1.2
- Reliability: Improved handling of long-running responses and clearer gateway timeout errors. Learn more
- Long tasks: When a request pauses at its turn limit, add turns and resume or stop. Apply an increased limit to the request, the chat, or the device default. Learn more
- Project conventions: Load project instructions from
AGENTS.mdand its referenced Markdown documents alongside project memory, refreshed for each request. Learn more - Models: Expanded model selection and aligned pricing tiers with configured model prices. Learn more
2.1.0
- Addressables: Inspect, configure, validate, build, and test Addressable content. Learn more
- Composer: Drag supported images, scripts, prefabs, ScriptableObjects, materials, scenes, shader sources and graphs, text and configuration files, project folders, and scene objects directly into a request. Learn more
- Models: Select Chat and Strategy models separately, with reasoning-effort controls for supported GPT-5.6 models. Learn more
- Change review: Clearer responses with View changes, Keep changes, Undo changes, and Auto-approve. Learn more
- 3D assets: Full-size previews, more predictable scene scale, scene-ready prefabs, and assisted placement. Learn more
2.0.0
- Change review: Review, keep, or undo project changes using persistent checkpoints. Learn more
- Asset generation: Create images, sprites, textures, 3D models, audio, and fonts inside Unity. Learn more
- Visual context: Attach images and capture Scene or Game views for analysis. Learn more
- Unity authoring: Work with prefabs, Animator Controllers, materials, shaders, lighting, particles, tags, and project settings.
- Verification: Check compilation, Play Mode behavior, scene layout, physics, and performance. Learn more
- Workspace: Use project memory, asset mentions, context usage, and per-request cost summaries.