Your plan, your inference
Forge runs the OpenAI Codex binary as a sidecar. You sign in with your ChatGPT account, and the agent uses your plan’s inference budget. There is no Forge proxy in the middle, and the model stays whatever ChatGPT’s latest is.
Reviewed before applied
Code edits land as diff cards with per-hunk accept and reject. Memory writes land as proposals you confirm. Forge never touches your files silently.
Engine-native bridges
Connect Unity for live console, hierarchy, and write tools. Connect Godot for scene-tree inspection and GDScript outline. Web stacks run inside a live preview tab.
Custom tools, not just chat
Asset studio for voice, music, sound effects, sprites, 3D, and motion capture. Pipeline tools for palettes, sprite sheets, tilemaps, 9-slice, shaders, Aseprite, and audio trimming. A task runner. Indexed memory search across your project.
