Run a capability, poll a task or fetch a result
Runs an inspected capability with your API key; a start spends balance at the
inspected price. Copy next from Inspect and fill the placeholders.
operation: start(default):inputfollows the inspectedinput_schema; unknown fields insideinputare rejected. Synchronous kinds — social and web data, text models, JEV — return the result in the response. Asynchronous kinds — image, video, workflows anddata:web.research(30 s to 3 min) — return{data: task, next}; repeatnext, anoperation: statuscall withtask_id, untilsucceededorfailed. A finished research task carries its result indata.output; a failed one is not charged.- Text models:
{"reference":"model:<id>","input":{"input":"<prompt>"}}returns{object:"text.result", model, status, output_text, usage, request_id}.input.inputis a string or a messages array;instructions,max_output_tokensandtemperatureare optional. - JEV (
model:jev-1.13,model:jev-1.13-free):{"input":{"state":…,"questions":…}}returns{id, model, answers, usage}. - Result views:
view: previewlifts the result’s main list toitems(the firstmax_items, default 10, each element trimmed inside) withitems_pathanditems_total, shortens long strings and marks the replytruncatedwith aresult_ref;fieldskeeps only the listed paths,items[].<key>for keys of each list element. REST defaults tofull; the BeatAPI MCP server defaults topreview. A status poll takes the same view. operation: resultwithrequest_idreturns a finished result again, free, for one hour, with anyview,fieldsormax_items.
Send a unique idempotency_key per task as a top-level field (or the
Idempotency-Key header) and reuse it only to retry the same start. An unknown reference returns 404 with
suggestions.
Authentication
Send your API key as Authorization: Bearer <BEATAPI_API_KEY>.
Headers
Dialect of next: http (default) writes a curl command, mcp writes a capabilities tool call. The BeatAPI MCP server sends mcp.
Request
start: run it. status: poll an async task (needs task_id; takes view, max_items and fields like start). result: fetch a finished result again, free within one hour (needs request_id).
For start: the input, following the inspected input_schema. Text models: {"input": "<prompt or messages>"} with optional instructions, max_output_tokens, temperature. JEV: {"state": …, "questions": …}.
For status: the task id an async start returned.
For result: the request_id a sync start returned.
full (REST default): the whole result. preview: the result’s main list lifted to items (the first max_items elements, each trimmed inside) with items_path and items_total, long strings shortened, truncated and result_ref reported.
With preview: elements kept in items (default 10).
Keep only these paths. Paths starting items[] are read from each element of the result’s list and returned as items, such as items[].title; other paths are dotted from the root, [] walks an array.
Unique per task, a top-level field next to reference and input (or the Idempotency-Key header); reuse only to retry the same start.
Response
Task status: poll until succeeded or failed.
Pass to operation: result to fetch the stored result.
With preview or items[] fields: the elements of the result’s main list.
Where the list sits in the full result, such as data.data.items or results.
Elements in items.

