Shorty

Shorty turns long videos, audio files, documents and web pages into readable summaries and accurate transcripts. The Shorty connector lets Claude work with that library from your own account: search your summaries, open a specific article and read its highlights and body, list what you made recently, list and open your transcriptions, check how much of your plan's quota is left, and search Shorty's developer documentation. It can also start new work - summarize a YouTube link, a web page, pasted text or a file already uploaded to your Shorty account; transcribe a publicly accessible audio or video URL; and add subtitle tracks to a video you provide. Claude reaches these through two tools, search_tools and execute_typescript.

These are long-running jobs, not instant answers. A write tool returns a job id, and Claude polls get_job_status until the job reaches SUCCESS, ERROR or CANCELLED before fetching the finished article or transcript. Over the connector Shorty returns job handles and text, never a media file, and subtitle files are downloaded from the Shorty app, not from the chat. Write tools are backed by an idempotency ledger, so a retried request will not create a duplicate job, and no tool here deletes or overwrites work you already have.

Access runs through OAuth 2.1 with PKCE. You choose which scopes to grant on the consent screen - separate read scopes for articles, transcriptions, jobs, usage and docs, and separate write scopes for creating summaries and transcriptions - and any tool whose scope you did not grant is never offered to Claude at all. Consent is re-checked on every individual tool call, so revoking Shorty from Settings > Connected AI Apps stops the assistant on its very next call, not at token expiry. Summaries and transcriptions draw on your Shorty plan's monthly quota, and Claude can read the balance with get_usage_quota before starting a long job. Nothing is sold in chat, no images, video or audio are generated, and no ads are served.