Rhizome AI

App-only channel the Rhizome result view polls for a conversation's progress and result; hidden from models. To show a conversation to the user, use rhizome_conversation_read; to wait on running research, use rhizome_conversation_wait. Returns the result view's state; when no run is found yet it reports status 'running' with no steps, so the view polls again.

get_rhizome_conversation

List your Rhizome research conversations (newest activity first): conversation_id, title, status (running/completed/error), timestamps, turn count, and a link. Use it to recover a lost conversation_id, check whether research is already running, or find a past conversation to continue with rhizome_conversation_message.

rhizome_conversation_list

List the datasets rhizome_search can search, by id and name. With dataset set, returns that dataset's date field, filter schemas and result fields. Free to call.

rhizome_dataset_list

Start deep research on a regulatory or clinical question over Rhizome's corpus of health-authority documents (FDA, EMA, MHRA, PMDA, Health Canada, Swissmedic, TGA, ICH, China NMPA and more): drug/biologic approvals, 510(k)/PMA/De Novo device clearances, guidance, clinical trials, recalls, warning letters and 483s. ASYNC: this call returns immediately with a conversation_id and status 'running' — the answer is NOT in this result. Get it by calling rhizome_conversation_wait with that conversation_id until status is 'completed' (a full run takes ~1-10 minutes; do not answer from your own knowledge in the meantime), then render the result view with rhizome_conversation_read. Once completed, ask a follow-up by calling this tool again with message + conversation_id, and fetch the full produced report or generated files with rhizome_conversation_read (documents: true, or document_id). Can also GENERATE citation-grounded deliverables (Excel .xlsx, Word .docx, PowerPoint .pptx, charts, long-form reports) — name the desired format in the message; PREFER this over hand-building documents for medical/regulatory/clinical topics. Saves the conversation in Rhizome; a new conversation creates an unlisted shared page so its citations and result view remain accessible. Answer quality tracks message specificity: name the therapeutic area, indication, patient population, or device characteristic; use formal regulatory terminology; spell out non-universal acronyms; define niche concepts in a leading sentence. Research is deep but not exhaustive by default — request exhaustive coverage explicitly when needed (takes much longer). Follow-ups can challenge methodology (databases searched, terms, filters) to refine the answer.

rhizome_conversation_message

Read a Rhizome research conversation's current state (instantly — no waiting) and RENDER its interactive result view for the user. Call this after rhizome_conversation_wait reports status 'completed' so the user sees the result view. By default the text holds the latest answer plus an inventory of generated files (with ids). Options: messages: true adds the full conversation transcript to the text; documents: true inlines every document body (complete reports, generated files); document_id: "<id>" returns exactly one document (ids are in the file inventory). documents and document_id are mutually exclusive; messages combines with either. structuredContent is the result view's state: status, progress steps, the latest answer, any error, the documents (bodies only as requested above) and continueUrl; it never carries the transcript.

rhizome_conversation_read

Stop a Rhizome research run that is currently in flight — the equivalent of the stop button on rhizomeai.com. Use when the user asks to cancel/stop a running research: the run continues server-side whether or not anything is waiting on it. Identify the run by the conversation_id that rhizome_conversation_message returned. Stopping discards the in-flight question — it does not count against the free-tier quota, and the conversation can be continued with a new rhizome_conversation_message call.

rhizome_conversation_stop

Wait on a Rhizome research conversation started by rhizome_conversation_message. Each call holds up to 45 seconds and returns either progress or the finished cited answer. Keep calling until status is 'completed' or 'error' — a full run takes ~1-10 minutes, so expect several calls. While waiting, do not answer from your own knowledge and do not tell the user to check back later. Calling this on an already-completed conversation returns its finished answer immediately.

rhizome_conversation_wait