WhisperAI

Search, transcribe, and summarize your recordings

WhisperAI connects your transcript library to your AI, so you can work with your recordings without leaving the conversation. Sign in with your whisperai.com account and your AI can search your transcripts, open any of them as plain text, timestamped segments, or SRT/VTT captions, and pull an AI summary of a completed transcript.

MCP connector

Server URL: https://whisperai.com/mcp
Login: None
Protocol: MCP 2025-11-25
Instructions: WhisperAI is an AI transcription service: users upload audio/video (meetings, interviews, lectures, calls) and get transcripts, speaker labels, AI summaries, and translations. When the user asks about recordings stored in WhisperAI, use to find the relevant recording before reading or changing it. To transcribe a local file from the user's device, call : it renders an in-conversation uploader widget. For audio/video already hosted at a publicly reachable URL, use instead. Both ingest tools accept optional transcription settings: language, transcriptionStyle (verbatim, clean_readable, domain presets), speakerNames or speakerRoles, speakerCount, importantTerms, customPrompt, and translateTo. Pass whichever the user has mentioned in the conversation. Anything unset falls back to the user's saved WhisperAI default preset and then to service defaults. Review those defaults and the requested settings with the user before they select a file. Upload is the final action and starts transcription immediately using normal plan minutes. The complete applied settings are shown in the uploader and returned to the agent. Call again to change them before upload; users can select up to 20 files together with the displayed settings, or use a separate uploader for files needing different settings. Authorization renews automatically, so the uploader has no 15-minute deadline. Track every recording ID reported by a batch. To change an existing recording, agree the revised settings and normal minutes usage with the user, then call . It reuses stored audio and preserves the original transcript. Reuse the same requestId and arguments on retries; poll the returned recordingId instead of submitting again. Transcription is asynchronous. After or , may return status "processing" — poll it about every 10 seconds until status is "completed". Transcripts can be very long. returns character-offset chunks: pass offset=nextOffset from the previous result to read more, and use the limit parameter to size chunks. returns the AI-generated summary (key points, action items) when one exists; reports the user's plan and remaining minutes. When the user wants a transcript as a downloadable file (PDF, Word, text, or SRT/VTT subtitles), call and show them the returned link.

Tools

Reads data6

Changes data5

Deletes or can't be undone1

Skills

None in the store listings or on the server

MCP connector

Server URL

https://whisperai.com/mcp

None

Protocol

MCP 2025-11-25

Instructions

WhisperAI is an AI transcription service: users upload audio/video (meetings, interviews, lectures, calls) and get transcripts, speaker labels, AI summaries, and translations. When the user asks about recordings stored in WhisperAI, use to find the relevant recording before reading or changing it. To transcribe a local file from the user's device, call : it renders an in-conversation uploader widget. For audio/video already hosted at a publicly reachable URL, use instead. Both ingest tools accept optional transcription settings: language, transcriptionStyle (verbatim, clean_readable, domain presets), speakerNames or speakerRoles, speakerCount, importantTerms, customPrompt, and translateTo. Pass whichever the user has mentioned in the conversation. Anything unset falls back to the user's saved WhisperAI default preset and then to service defaults. Review those defaults and the requested settings with the user before they select a file. Upload is the final action and starts transcription immediately using normal plan minutes. The complete applied settings are shown in the uploader and returned to the agent. Call again to change them before upload; users can select up to 20 files together with the displayed settings, or use a separate uploader for files needing different settings. Authorization renews automatically, so the uploader has no 15-minute deadline. Track every recording ID reported by a batch. To change an existing recording, agree the revised settings and normal minutes usage with the user, then call . It reuses stored audio and preserves the original transcript. Reuse the same requestId and arguments on retries; poll the returned recordingId instead of submitting again. Transcription is asynchronous. After or , may return status "processing" — poll it about every 10 seconds until status is "completed". Transcripts can be very long. returns character-offset chunks: pass offset=nextOffset from the previous result to read more, and use the limit parameter to size chunks. returns the AI-generated summary (key points, action items) when one exists; reports the user's plan and remaining minutes. When the user wants a transcript as a downloadable file (PDF, Word, text, or SRT/VTT subtitles), call and show them the returned link.