Prompt has no outputs: provide an actual execution endpoint

Symptoms and scope

A graph of loaders and intermediate calculations does not necessarily specify an output to execute. Locate a valid result endpoint rather than the visually rightmost node.

Error fragments for search; IDs, values and filenames may vary:

Prompt has no outputs
prompt_no_outputs

Source-supported context

Core validate_prompt identifies nodes declared OUTPUT_NODE and considers partial-execution targets. The official basic workflow ends its decoded image path at a save node. Source 1 Source 2

Cases to distinguish

1. Exporting or manually trimming the request removed save, preview or another valid output node.

2. partial_execution_targets is set but contains no ID of a registered OUTPUT_NODE in the submitted graph.

3. The apparent endpoint is an intermediate node rather than a backend class declared OUTPUT_NODE=True. A missing custom node class is a different missing_node_type error; a present output with invalid upstream inputs is prompt_outputs_failed_validation.

Diagnostic sequence

The sequence below is an editorial procedure based on the cited context, not a diagnosis already confirmed for your environment.

Step 1. Identify the node that actually saves or returns the required result in this workflow.

Step 2. Inspect the submitted API graph, not only the canvas. Confirm it includes the endpoint and that the installed backend class marks it as an output. If the server instead reports missing_node_type, fix that missing class first; if it reports prompt_outputs_failed_validation, inspect the reported node errors and upstream inputs.

Step 3. If sending partial_execution_targets, match those IDs to output nodes in the current API graph. Retry the complete native API export without a partial-target filter to isolate a stale or over-narrow filter.

Step 4. If a custom endpoint is intended, verify that its extension is installed and its server-side class declares OUTPUT_NODE=True. Adding an OUTPUT_NODE field to the submitted request cannot change the registered class.

Completion check

The /prompt response supplies prompt_id only after validation accepts the request. Track that ID through execution and inspect its history/output result; acceptance alone does not prove a file was saved.

Limitations and cautions

Queue acceptance is not file generation. Output can be a preview or another node-defined result, not always a file in one output directory.

Original sources

• ComfyUI execution.py validation — checked 2026-09-26.
• ComfyUI server.py prompt handling — checked 2026-09-26.
• Text-to-image workflow — checked 2026-09-26.

Source and English-language review: 2026-09-26. The diagnostic branches above follow current ComfyUI validation code; no ComfyUI workflow, file save or GPU run was performed by this site.

Related troubleshooting and guides

• Queue finishes quickly but no new file appears: check cache and output location
• Submit a ComfyUI workflow through its native API and track prompt_id