Sonix

Generate a downloadable Sonix transcript in the requested format and return a short-lived (30 minute) URL. The transcript content does NOT enter conversation context: the URL is meant to be opened by the user, or downloaded by a client that has file access. Use this for verbatim file output, captioning files, JSON for downstream processing, or transcripts too large for `read_transcript`. Supported formats: txt, srt, vtt, json. Format-specific options (speaker labels, timestamps, subtitle line splits, etc.) map directly to the existing Sonix REST API export options. Recordings still transcribing return a structured error.

export_transcript

Return the authenticated user's Sonix account: plan and current credit balance broken out by source (free, paid, included). Use this for "how much credit do I have?" or to gate other actions on remaining balance. Credit values are in seconds; divide by 3600 for hours. The three sources always sum to total.

get_account_info

List media (audio/video recordings) belonging to the authenticated Sonix account, newest first, 25 per page. Returns metadata only — use read_transcript to fetch transcript content for a specific item. Use this when the user asks "what recordings do I have?" or wants to find a recording to act on. Returns hashid IDs (like "Ko2dyxm4") that other tools accept.

list_media

Pull a Sonix transcript into the conversation context for analysis, Q&A, summarization, sentiment, entity extraction, or any task where the LLM should reason about the transcript content. Returns speaker-labeled text — the shape the model reasons over best. The transcript counts against your context window. For verbatim file output (txt/srt/vtt/json), captioning, or transcripts too large for inline reading, use `export_transcript` instead. Recordings still transcribing return a structured error — try again once status is `completed`.

read_transcript