CoreWise

Start an AI extraction of a YouTube video, podcast, article, or PDF URL on CoreWise. Returns an extraction_id immediately after initialization. Initialization normally takes a few seconds but can take up to 2 minutes for videos without captions or for PDFs. The extraction itself then runs for 1-7 minutes: poll with get_extraction every 20-30 seconds until status is 'completed'. Results include a cross-validated synthesis plus per-model summaries. Requires an API key (create one at corewise.video, Profile page, 'API & MCP Keys'). Each call consumes one extraction from the key owner's monthly quota.

{
  "type": "object",
  "properties": {
    "url": {
      "type": "string",
      "description": "The video, podcast, article, or PDF URL to extract insights from."
    },
    "models": {
      "type": "array",
      "items": {
        "type": "string",
        "enum": [
          "claude",
          "openai",
          "gemini",
          "mercury",
          "grok",
          "mistral"
        ]
      },
      "description": "Which AI models to run (1-6). Default: ['openai', 'gemini']. More models cost more time."
    }
  },
  "required": [
    "url"
  ]
}

Get the status and results of a CoreWise extraction by id. While status is 'processing', poll every 20-30 seconds. When status is 'completed' the response contains the structured synthesis (JSON object: summary, takeaways, highlights, notable quotes) and per-model markdown summaries. Works on your own extractions (API key required) and on public community extractions.

{
  "type": "object",
  "properties": {
    "extraction_id": {
      "type": "string",
      "description": "The extraction id returned by extract_url."
    }
  },
  "required": [
    "extraction_id"
  ]
}

Get the raw source transcript (or article/PDF text) of a CoreWise extraction by id. Available as soon as the extraction has been initialized — no need to wait for 'completed'. Transcripts can be very long, so results are paged: pass 'offset' and 'max_chars' to fetch further chunks when 'truncated' is true. Works on your own extractions (API key required) and on public community extractions.

{
  "type": "object",
  "properties": {
    "extraction_id": {
      "type": "string",
      "description": "The extraction id returned by extract_url or found via search."
    },
    "offset": {
      "type": "number",
      "description": "Character offset to start from (default 0). Use with max_chars to page."
    },
    "max_chars": {
      "type": "number",
      "description": "Max characters to return (default 100000, max 200000)."
    }
  },
  "required": [
    "extraction_id"
  ]
}

Search CoreWise's public community library of extractions (videos, podcasts, articles, and PDFs already analyzed by other users). Returns titles, snippets, and share URLs. Useful to check whether a video has already been extracted before starting a new extraction. No API key required.

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "Search terms matched against titles and content."
    },
    "limit": {
      "type": "number",
      "description": "Max results (default 10, max 25)."
    }
  },
  "required": [
    "query"
  ]
}