Gallery and assets

The gallery and asset tools let an assistant find, show and reuse what you already made in Nodaro. browse_gallery and browse_uploads show your results and your uploads as a grid, get_asset and display_asset fetch or show one result, get_app_run follows a workflow or app run, and list_favorites and favorite_asset manage your favorites. None of them costs credits.

browse_gallery

Shows your gallery as an interactive grid of thumbnails. You can click a tile to view it, copy it, or use it in the next edit. Use it when the user wants to see their work; use list_jobs when the assistant only needs data.

Permission: assets:read. Credits: free.

Returns: the grid in clients that display MCP Apps, and the items with a next_cursor as data.

The public scope shows what the public gallery on the web shows, with the same moderation. A page can therefore hold fewer items than limit: keep following next_cursor until it is absent.

On a deployment that does not allow public work (outputs.allowPublic: false), the public scope lists nothing and answers (no items). Your own gallery does not change.

Your own gallery also lists your Apply EDL renders: a video render under video, and an audio render under audio. A render made at Preview quality carries preview: true, and its text line says (preview). The public scope never lists an Apply EDL render.

browse_uploads

Shows the files you uploaded, as opposed to the results Nodaro generated. Use it to pick an earlier upload as the input of a new generation. The grid's Use button hands the file's id back to the chat.

Permission: assets:read. Credits: free.

Returns: the uploads with their URLs and ids, and a nextCursor. An Apply EDL render made at Preview quality carries preview: true, and its text line says (preview). A file whose stored copy is gone is not listed and not counted.

get_asset

Returns the details of one result: its output URL, prompt and model. It returns your own jobs in any status, and the public, completed jobs of other users. On a deployment that does not allow public work (outputs.allowPublic: false), it returns your own jobs only, and another user's job reads as not found. A public job of another user carries no provider words for a failure (errorDetail is null) and not that user's id. For a failed job, it adds retryable, guidance and, when available, suggestedProvider; When a job fails explains them.

Permission: assets:read. Credits: free.

Returns: the job envelope, without input.

display_asset

Shows one result in the chat: the user sees the image, video or audio instead of data. Images also get Animate, Edit and Recreate buttons and a full-screen view. Use get_asset when the assistant only needs the details.

Permission: assets:read. Credits: free.

On a deployment that does not allow public work (outputs.allowPublic: false), display_asset shows your own jobs only. Another user's job reads as not found.

Returns: the media card, with the result's details as data.

get_app_run

Returns the progress of a workflow run or a published app run: its status, the state of each node, and every output URL produced so far, with its prompt, model and job id. Use it after run_workflow, render_final and run_app, which return a run id (execution_id) instead of a job id. get_job does not find a run id. Call it every 5 to 10 seconds while the run is pending or running.

Permission: assets:read. Credits: free.

Returns: the run's status, the node states and the outputs so far:

• executionId, status, errorMessage, and a summary with the counts total, completed, failed, skipped and skippedForEmptyInput.
• nodeStates, one entry per node in the workflow's order: id, label, nodeType, status, error, jobId, text and media. text is the node's text output, cut at 1,500 characters, with textTruncated: true when it was cut. media lists every image, video and audio URL the node produced, each as { kind, url }. skipReason: "empty_input" means nothing reached the node's input in this run.
• outputs, every media URL of the run, with the prompt, model and createdAt of the job that made it.
• resumeHint, on a failed run that run_workflow can continue (a saved workflow in the mcp project, not an app run): the run_workflow call that continues it from where it stopped. See Continue a failed run. A failed app run carries the app's Retry instead: the retry_app_run call that quotes it. Only the run's current execution carries it, so after a Retry or a fresh start, the earlier execution gets none. The call names the run's own version when the app changed since the run.

When the run completes, it also carries outcome. The outcome is "nothing_new" when at least one node was skipped because nothing reached its input and nothing failed, for example a feed that found no new posts. Otherwise it is "succeeded". A failed or cancelled run has a status, not an outcome.

list_favorites

Lists the job ids of your favorite results, newest first.

Permission: assets:read. Credits: free.

Returns: the job ids of your favorites. A favorite Apply EDL render of your own is listed too, with preview: true on a Preview. Another person's render is not. On a deployment that does not allow public work (outputs.allowPublic: false), only your own jobs are returned.

favorite_asset

Adds a result to your favorites or removes it. The result can be one of your own jobs or a public job of another user. On a deployment that does not allow public work (outputs.allowPublic: false), you can mark only your own jobs. Removing a favorite always works. list_jobs with favorited: true lists the jobs you favorited.

Permission: assets:write. Credits: free.

Returns: a confirmation.