Speak AI

Ask an AI-powered question about your media using Speak AI's AI Chat. Supports querying a single file, multiple files, entire folders, or your whole workspace. Pass mediaIds for specific files, folderIds for entire folders, or omit both to search across all media. Use assistantType to get specialized responses (e.g., 'researcher' for academic analysis, 'sales' for deal insights). To continue a conversation, pass the promptId from a previous response. Returns a promptId — save it to continue the conversation with follow-up questions.

High-level automation builder: create (or update) a Speak automation from a friendly spec without knowing the wire format. Accepts folder/custom-field NAMES (resolved to ids; missing folders are auto-created), payload.<path> shorthand for webhook tokens, and simple step types (filter, branch, upload, ai_chat, translate, notify, call_webhook). For inbound-webhook automations the result includes the receive URL and mappable payload tokens. Prefer this over create_automation unless you need raw control.

Set the folder scope for multiple automations at once. Pass an empty folderIds array to remove the folder restriction (run on all folders).

Move multiple media files to a folder in a single operation. Use this for batch reorganization instead of updating media one by one.

Activate or deactivate multiple automations at once.

Normalise speaker names that are ALREADY correct across multiple media files — for example changing "Frederik S." to "Frederik" everywhere. Applies the same mapping to every specified media file. NOT a way to identify a speaker across a project: neither the numeric id nor a default label such as "Speaker 1" refers to the same person in different files, because speakers are numbered per file in order of appearance. Renaming "Speaker 1" across many files will label a different person in each one. Identify the speakers in each file first (get_transcript, or an identify-speakers automation), then use this tool only to tidy up naming that is already correct. Match by the speaker's current LABEL, not by numeric id. A file whose speakers do not match the mapping is left unchanged and reported as failed. Re-sending a mapping that has already been applied is a safe no-op, so do not retry a call that reported success.

Check if an embed exists for a media file and retrieve its configuration.

Check whether a recorder/survey is active and accepting submissions.

Duplicate an existing folder and all of its contents.

Duplicate an existing folder view into a target folder.

Duplicate an existing recorder including all its settings and questions.

Create a new automation rule using the V2 graph model (trigger + ordered steps). Fetch valid step/trigger options with list_automation_triggers / list_automation_actions if unsure. For inbound-webhook automations the response includes inboundWebhook.inboundUrl (where to POST payloads) — recommended flow: create, send a test payload to the URL with ?test=1, call get_inbound_webhook to see mappable payload tokens, then update_automation to wire tokens/fieldsMap.

Create a highlight clip from one or more media files by specifying time ranges. Clips are processed asynchronously (states: queued, processing, completed, failed) — use get_clips to check status. Maximum total clip duration is 30 minutes. Use multiple timeRanges to stitch segments from different media files together.

Create an analytics dashboard. Only `title` is required — source defaults to the whole workspace and dateRange to last30days. Add widgets by listing their types (the MCP assigns ids and lays them out automatically), scope with source ({type:"folders",folderIds} | {type:"team"} | {type:"workspace"}) and dateRange ({preset}), and optionally group widgets into sections. Design guidance: lead with a narrative widget as the first widget; group sections by the QUESTION they answer, not by widget type; don't pad — every widget earns its place (aim for 4-16 widgets on a full build); if something can't be expressed by the widget catalog, put it in a narrative widget's focus instead of faking it. Call list_dashboard_widgets first for the widget catalog, config vocabulary, design rules, and full examples.

Create an embeddable player/transcript widget for a media file or a set of folders. Provide `mediaId` for a single-media embed, or `folderIds` for a folder/library embed.

Create a new custom field for categorizing and tagging media.

Create a new folder in the workspace.

Create a new saved view for a folder with a custom set of display columns.

Create a new recorder or survey for collecting audio/video submissions.

Create a new text note in Speak AI for analysis. The content will be analyzed for insights, topics, and sentiment.

Create a new user group and assign members. Member ids come from list_users. Fails with a 409 if a group with the same name already exists in the company.

Create a new webhook to receive real-time notifications when events occur in Speak AI.

Permanently delete an automation rule.

Delete a specific chat message from conversation history.

Permanently delete a clip and its associated media file.

Soft-delete a dashboard. This also deactivates its public share link.

Permanently delete a folder. Media within the folder will be moved, not deleted.

Permanently delete a media file and all associated transcripts and insights.

Permanently delete a recorder/survey. Existing recordings are preserved.

Cancel and delete a scheduled meeting assistant event.

Delete a user group. This removes the group only; it does not delete the users themselves.

Delete a webhook and stop receiving notifications at its endpoint.

Clone an existing dashboard. The copy gets fresh widget ids, a "<name> (copy)" title, cleared sharing, and its revision reset to 0. Ideal for cloning a fully-configured dashboard, then tweaking it via update_dashboard.

Export a specific AI Chat answer. Useful for saving AI-generated summaries, reports, or analysis results.

Export a media file's transcript or insights in various formats (pdf, docx, srt, vtt, txt, csv).

Export multiple media files at once, optionally merged into a single file.

Generate a shareable public URL for a recorder/survey.

Retrieve all saved views across all folders.

Get detailed information about a specific automation rule, including its trigger and step graph.

Get the run history (executions) for an automation, with paging and optional status filter.

Get captions for a media file. Captions are separate from full transcripts and are formatted for display/subtitles.

Get a list of recent AI Chat conversations. Returns conversation summaries with promptIds that can be used to continue conversations via ask_ai_chat or retrieve full messages via get_chat_messages.

Get full message history for conversations. Can filter by promptId for a specific conversation, by media/folder, or search across all chat messages. Returns questions, answers, references, and metadata.

Get usage statistics for AI Chat / chat. Returns metrics on prompt usage, optionally filtered by date range.

List clips, optionally filtered by folder or media files. If clipId is provided, returns a single clip with its download URL (when processed).

Get a single dashboard's full spec: title, description, source, date range, sections, widgets, and the current `revision` (pass that revision back to update_dashboard).

Compute a speakers breakdown for a given folder scope, date range, and field filters. Standalone analytics — does not require a dashboard to exist.

Get the iframe URL for embedding a media player/transcript on a webpage.

Get all prompts and answers that have been marked as favorites. Useful for finding saved insights and important AI-generated analysis.

Get detailed information about a specific folder including its contents.