list_fan_out_extractions
list_fan_out_extractions MCP tool: the paid fan-out extractions of a project, newest first, with status, prompt count, credits charged and progress. Filters, pagination, JSON example.
Updated 2026-09-26
list_fan_out_extractions lists the fan-out extractions of a project, newest first. An extraction queries ChatGPT's search engine through the OpenAI API for chosen prompts and stores its queries, results and citations (see create_fan_out_extraction). Each row carries the status, the number of prompts, the credits actually charged and the progress prompt by prompt, so an agent can tell what is finished without opening each one.
When to use
- Before launching a new extraction, to reuse a recent one on the same prompts.
- To find the
extractionIdto pass toget_fan_out_extraction. - To follow the extractions still running (
filters.status: ["PENDING", "RUNNING"]).
Input
| Field | Type | Default | Description |
|---|---|---|---|
projectId |
string (CUID) | required | Project scope. |
filters.status |
enum[] | all | PENDING, RUNNING, COMPLETED, PARTIAL, FAILED. Omit to return every status. |
cursor |
string | - | nextCursor from the previous page, echoed back as is. |
limit |
integer | 20 |
1 to 100. |
Response
{
"data": [
{
"id": "clx_ext_42",
"status": "COMPLETED",
"llm": "CHATGPT",
"origin": "app",
"promptCount": 24,
"creditsCharged": 110,
"progress": { "pending": 0, "done": 22, "empty": 2, "failed": 0 },
"createdAt": "2026-09-26T10:00:00.000Z",
"completedAt": "2026-09-26T10:02:10.000Z"
}
],
"pageInfo": { "hasMore": false, "nextCursor": null, "totalCount": 1 }
}
status is PARTIAL when some prompts failed for good (they are not charged), FAILED when all of them did. origin tells whether the extraction was launched from the app or through the MCP. creditsCharged counts only the prompts where the engine searched.
Tips and patterns
- Running first. Filter on
PENDINGandRUNNINGto poll only what is still moving. - Progress without the payload.
progressgives done, empty, failed and pending counts; open the detail only when needed. - Compare over time. Two extractions on the same prompts, weeks apart, show how the engine's searches and results moved.
Related tools
- get_fan_out_extraction: read one extraction.
- create_fan_out_extraction: launch one.
- list_fan_outs: fan-out queries captured by daily tracking.