MEASURED SCOPE / 实测范围

Evidence you can inspect.

Run a workflow Plan a task Browse 26 candidates Inspect task tests My toolbox Advanced API guide

Report generated: 2026-10-22T07:59:06.487Z. Named assertions on public research calls and generated English PDF fixtures. Failing checks remain visible. No general benchmark ranking, customer success rate or scanned-document accuracy claim.

arXiv API: connection plan

none · https-api · Setup source

https://export.arxiv.org/api/query

Documentation checked: 2026-10-22 · Pricing: documented_free_access_or_local_software

Supported operations and exact formats
{
  "scholarly_research": {
    "inputs": [
      "text"
    ],
    "outputs": [
      "sources",
      "structured_data"
    ]
  }
}
  1. Read current provider setup, allowed data handling and pricing.
  2. Parse Atom XML, not JSON. Serialize requests and follow the provider rate guidance.
  3. Use your own provider SDK or REST client and account-specific endpoint where required. Credentials remain with your client.
  4. Try a small authorized task. Evaluate the output against your requirements.
  5. Record a private attempt in your workspace. Publish feedback only after reviewing it.
{
  "tool": "GET /api/query",
  "arguments": {
    "search_query": "all:document AND all:extraction",
    "start": 0,
    "max_results": 5
  },
  "placeholder_values_require_replacement": true,
  "validate_current_provider_schema": true
}

Preprints are not necessarily peer reviewed. The metadata API does not perform PDF extraction.

Machine-readable connection and retry rules

Dated task assertions

Only the assertion below was checked. Local PDF tests use generated English documents. These results do not measure customer success, OCR accuracy or future availability. Tests older than 30 days do not count as a recent pass in task planning.

Passed · arxiv-atom-metadata-v1

The public query returns an Atom entry with a title and arXiv identifier.

2026-10-22T07:59:06.486Z · public_read_only_task · 1066 ms (single check)

One query checks metadata fields only. No full-text extraction, source quality, general relevance or reliability claim.

svc_official_arxiv connection plan

JSON test report

Repeat with your own criteria

SDK end-to-end checks: public research, local field provenance, unsupported scans and cancellation · Run the workflows

Run the local PDF fixtures with npm test in the downloadable example. Research test queries and tools are included in the JSON report. Initialize the provider and inspect tools/list before repeating them. No test runs automatically against your files.

Download the task planning client