- Lay out first, generate later: an AI tool can add image, video, audio or text nodes to your personal canvas with the prompt, model, settings and reference images ready — nothing is generated and no Xins are spent. You review and adjust them in the canvas, then press generate yourself. Use
xinyu_create_nodein MCP orxinyu node createin the CLI (requires MCP 0.1.52 / CLI 0.1.53 or later). - Your uploads no longer need confirmation: when an AI tool connects or re-configures an image you uploaded, it is no longer treated as a paid result that needs extra confirmation.
- Find and inspect nodes from the CLI: new
xinyu node find(search by title, prompt and more),xinyu node inspect(full details) andxinyu node read-prompt(the full prompt). - Clean up empty canvases: once a test canvas an AI tool created is empty, it can be deleted with
xinyu_delete_projectorxinyu project delete. Only empty canvases can be deleted this way; delete anything else in the web canvas.
ChangelogNov 9, 2026