check_attached_document
ChatGPTNever call this to answer a capability question such as 'which checks can you run on this?' or 'what can Paperpal check' — answer that directly instead, even with a document attached; see instructions. Use when the user attaches a document directly to their message and asks for a Paperpal check on it. Uploads a remotely downloadable attachment to Paperpal and starts the requested checks in one step. Call this instead of show_document_upload when the host supplies both file_id and a real download_url. If the host supplies only a private file id or sandbox path, call show_document_upload; never invent a download_url. If the attachment is not a DOCX or PDF (for example a spreadsheet or image), this tool opens the portable Paperpal upload card instead of starting a run; do not call show_document_upload again in that case — just tell the user the card is ready. If the request names specific checks (for example 'run a plagiarism check'), pass those as checks; omit checks for an ambiguous request such as 'check my document', which runs every available check. Grammar is DOCX-only: instead of a polled result, it hands the document to the Paperpal editor (grammar runs automatically there) and returns an editor link; it is not available for PDF — pass 'language' here even for a PDF whenever other checks are also requested, the tool drops it automatically and explains why in its reply and in the result's blocked_checks (see server instructions (1a) for the Grammar-only-on-PDF case, where this tool should not be called at all). Likewise, 'references' is automatically dropped and explained the same way for a free-plan account; never pre-filter checks for either case yourself. If blocked_checks comes back non-empty — including when the request was standalone and every requested check was blocked, so checks comes back empty — name every entry and its reason in your reply; never describe that instead as the run "completing with nothing found" (see server instructions (5)). This already returns a live progress card for the new check_run_id — do not immediately call get_check_run_status afterward in the same turn, that only duplicates the card just shown.
get_check_run_findings
ChatGPTFetch the findings for an existing check_run_id, as plain data with no UI card — use this to answer a follow-up question about results already shown in this conversation, such as 'help me fix the plagiarism issues' or 'what references need fixing'. Always pass the check_run_id already returned by start_document_checks, check_attached_document, or get_check_run_status earlier in this conversation. Do not call start_document_checks or check_attached_document again just to answer a follow-up — only call them if the user explicitly asks to rerun or restart checks.
get_check_run_status
ChatGPTFetch the latest live Paperpal status for a check run that already has a check_run_id. Use for requests to refresh, view, or show live progress for a manuscript check — including a hand-off message sent right after checks were started, such as 'Show me the live Paperpal check progress for "<filename>" (check run <uuid>)'. Always call this instead of start_document_checks or check_attached_document when a check_run_id is already known; this never starts a new run. Do NOT call this in the same turn immediately after start_document_checks or check_attached_document — their own result already renders a live progress card for that check_run_id, so an immediate follow-up call here only produces a confusing duplicate card. Call it only in response to a separate, later user message. Do NOT use this to explain what a completed check found, or to help fix an issue it reported — that re-renders the same card and answers nothing new. Call get_check_run_findings for that instead, every time the user is asking about or acting on results already shown in this conversation.
record_check_feedback
ChatGPTInternal app bridge for recording a "was this helpful" rating (and, for negative ratings, a reason and optional comment) against a completed check run.
record_product_event
ChatGPTInternal app bridge for recording allowlisted Paperpal result-card interactions.
show_document_upload
ChatGPTShow the Paperpal card for selecting or uploading a manuscript and then choosing checks. This is the default for generic requests such as 'check my paper' or 'is the manuscript ready to submit?' when no file is attached and no concrete existing Paperpal document is identified. If the current message has a DOCX or PDF attached and the host can supply a real download_url, call check_attached_document instead, even for a generic phrasing like 'check my paper' or 'help me run a reference check'. If the host exposes only a private file id or sandbox path, use this upload card; never invent a download_url. check_attached_document uploads and starts checks in one step and enforces plan/quota rules before ever showing a card, which this card's own picker does not. Do not call show_document_check_results for those generic requests. If the user gives a save-as name, pass document_name. If the request names specific checks, pass checks; otherwise omit checks so all available checks are pre-selected.
start_document_checks
ChatGPTStart the selected Paperpal checks (plagiarism, reference checks, journal fit, grammar) for an uploaded document. Use when the user asks to check, review, or assess readiness of their manuscript. Grammar is DOCX-only: instead of a polled result, it hands the document to the Paperpal editor (grammar runs automatically there) and returns an editor link; it is not available for PDF, and is automatically dropped (with an explanation in the reply, and in the result's blocked_checks) rather than shown as a broken result if requested for one. References is dropped the same way for a free-plan account. If blocked_checks comes back non-empty, name every entry and its reason in your reply — see server instructions (5).
upload_document
ChatGPTUpload a selected manuscript to Paperpal. Use after the user selects a DOCX or PDF file in the Paperpal card.
check_attached_document
Claudeget_check_run_findings
Claudeget_check_run_status
Claudeshow_document_upload
Claude
