plugin.json

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "app-6a094509363081918cb9dd23801d8f13",
  "version": "1.41.1",
  "description": "Capture meetings, search thousands of recordings, run async voice and video surveys, create clips, and automate workflows with Speak AI. 189 MCP tools, 5 resources, 3 prompts, and 8 skills.",
  "author": {
    "name": "Speak AI",
    "email": "accounts@speakai.co",
    "url": "https://speakai.co"
  },
  "homepage": "https://docs.speakai.co/mcp/plugin",
  "repository": "https://github.com/speakai/speakai-mcp",
  "license": "MIT",
  "keywords": [
    "mcp",
    "speak-ai",
    "transcription",
    "media",
    "insights",
    "meetings",
    "research",
    "clips",
    "captions",
    "automations"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Speak AI",
        "shortDescription": "Meeting and recording insights",
        "longDescription": "Speak AI transcribes and analyzes your interviews, sales calls, research sessions, webinars, podcasts, and team meetings. This connector brings that library into your AI assistant, so instead of scrubbing hours of audio, you just ask.\n\nCapture: Send a meeting assistant to Zoom, Google Meet, Microsoft Teams, or Webex. Upload audio, video, or text. Run async voice and video surveys with custom questions.\n\nAnalyze: Search transcripts and insights across thousands of recordings at once, with citations. Pull decisions, action items, owners, risks, sentiment, themes, and exact quotes.\n\nAct: Create clips, highlights, captions, notes, folders, fields, and embeds from chat. Chain steps with automations and webhooks: auto-tag new recordings, route insights to other tools, or run a custom prompt against every new upload.\n\nShare and organize: Export answers, transcripts, captions, and clips. Generate embed links and shareable views. Bulk move media, update transcripts and speakers, and clone setups across teams.\n\nWho it's for: researchers, revenue and customer-success teams, meeting-heavy operators, and media and content workflows: anyone who records conversations and wants to stop re-listening to find what matters.\n\nYour recordings stay in your Speak AI workspace. The connector only queries them with the permissions you allow.\n\nLimitations: the connector works only with recordings in the Speak AI workspace you connect. Deleting, overwriting, and anything that sends data outside Speak AI are marked so your assistant can check with you before running them.",
        "developerName": "Speak AI",
        "category": "Productivity",
        "capabilities": [
          "Read",
          "Write"
        ],
        "websiteURL": "https://docs.speakai.co/mcp/",
        "supportURL": "https://docs.speakai.co/help/",
        "privacyPolicyURL": "https://speakai.co/privacy-policy/",
        "termsOfServiceURL": "https://speakai.co/terms-of-service/",
        "defaultPrompt": [
          "Summarize this week's meetings into decisions and action items.",
          "Find customer quotes about onboarding, with the source recordings.",
          "Clip the best 30 seconds of the latest webinar and export captions."
        ],
        "brandColor": "#2563EB",
        "composerIcon": "./assets/composer-icon.png",
        "composerIconDark": "./assets/composer-icon-dark.png",
        "logo": "./assets/logo.png",
        "logoDark": "./assets/logo-dark.png"
      },
      "onboardingSkill": "./skills/getting-started/SKILL.md",
      "review": {
        "commerce": false,
        "commerce_description": "Speak AI does not sell products or process payments through this plugin.",
        "test_cases": {
          "positive": [
            {
              "description": "List recent media in the Speak AI workspace.",
              "prompt": "Use Speak AI to list my 10 most recent media files with each file name, media ID, media type, processing state, and created date.",
              "tools_triggered": "list_media",
              "expected_behavior": "Returns up to 10 recent Speak AI media records with names, media IDs, media types, processing states, and created dates, or clearly states that no matching media was found."
            },
            {
              "description": "Search recordings and transcripts for pricing feedback.",
              "prompt": "Use Speak AI to search my media library for recordings or transcripts that mention pricing feedback. Return the most relevant matches with media IDs and short snippets or summaries.",
              "tools_triggered": "search_media",
              "expected_behavior": "Returns relevant Speak AI search results for pricing feedback with enough source detail, such as media IDs, names, snippets, or summaries, for the user to open the recordings, or clearly states that no matching media was found."
            },
            {
              "description": "Check processing status for the most recent media file.",
              "prompt": "Use Speak AI to find my most recent media file and check its processing status.",
              "tools_triggered": "list_media, get_media_status",
              "expected_behavior": "Lists the most recent Speak AI media record, then returns its current processing state, such as pending, transcribing, analyzing, processed, or failed, for that media ID, and does not invent transcript content. If the workspace has no media, clearly states that."
            },
            {
              "description": "Retrieve a transcript for the most recent processed audio or video recording.",
              "prompt": "Use Speak AI to list my audio and video recordings, not text notes, find the most recent one that has finished processing, and get its transcript. Include speaker labels and timestamps if Speak AI returns them.",
              "tools_triggered": "list_media, get_transcript",
              "expected_behavior": "Identifies a processed audio or video recording from the workspace (not a text note), then returns its transcript data with speaker and timestamp information when available, or reports the exact Speak AI processing/error state if the transcript is not available. If no processed audio or video recording exists, clearly states that."
            },
            {
              "description": "Summarize decisions and action items from the most recent recording.",
              "prompt": "Use Speak AI to find my most recent processed audio or video recording, not a text note, and summarize it: list the key decisions, action items with owners, risks, and two notable quotes.",
              "tools_triggered": "list_media, get_media_insights, get_transcript",
              "expected_behavior": "Finds the most recent processed audio or video recording with list_media, reads its insights and transcript, and returns decisions, action items with owners, risks, and two verbatim quotes, naming the recording. It does not invent content that is not in the insights or transcript."
            }
          ],
          "negative": [
            {
              "description": "Do not invoke Speak AI for general web news.",
              "prompt": "Search the web for the latest OpenAI news and summarize the top stories.",
              "expected_behavior": "Speak AI is not called. The request is a general web search and has nothing to do with the user's Speak AI workspace."
            },
            {
              "description": "Do not invoke Speak AI for calendar scheduling without meeting-assistant intent.",
              "prompt": "Put a team lunch on my Google Calendar tomorrow at noon.",
              "expected_behavior": "Speak AI is not called. The request is ordinary calendar scheduling and does not ask for the Speak AI meeting assistant."
            },
            {
              "description": "Do not invoke Speak AI for text already in the conversation.",
              "prompt": "Summarize this paragraph in two sentences: Our team met on Monday to plan the launch. We agreed to ship the beta in two weeks and to collect feedback from ten customers.",
              "expected_behavior": "The Speak AI app should not be invoked because the text is already in the conversation and nothing needs to be read from or saved to the user's Speak AI workspace."
            }
          ]
        },
        "demo_recording_url": "https://drive.google.com/file/d/1BOqI-OdeqII2QQkUolQFVeHqycgbuoBB/view?usp=sharing"
      },
      "publication": {
        "release_notes": "Every operation the model can call is now its own tool with a description, input schema and annotations. AI Chat, which passed requests to Speak AI's own assistant and could call other tools, is no longer part of the hosted server, and the skills read transcripts and insights directly instead. Listing icons and logos now use the Speak AI wordmark, and REST calls authenticate with the API key alone."
      },
      "apps": "./.app.json"
    }
  }
}