Supreme Court oral arguments as vCons

Every argument with public audio, 1955 to 2025, as IETF vCon records with speaker-attributed transcripts. Read-only, free, no signup.

Endpoint https://mcp-scotus.vconic.com/mcp
Token Bearer scotus

8,503 arguments8,600 hours4.5M transcript segments

What a vCon is

A vCon is one conversation as a single JSON record, the IETF's container format for recorded conversations. Every record here is one argument session. It has four parts: parties (who was in the room: the justices and the advocates), dialog (the recording itself, here a link to the Court's MP3), analysis (things derived from the dialog, here the transcript, one segment per speaker turn with speaker pointing into parties), and attachments (documents about the conversation: the case metadata with docket, question presented and outcome, plus tags and the lawful basis for holding the recording). Trimmed example:

one vCon, abridged
{
  "subject": "Flast v. Cohen (416) - Oral Argument - March 12, 1968",
  "parties": [
    { "name": "Earl Warren", "role": "justice" },
    { "name": "Leo Pfeffer", "role": "advocate" }
  ],
  "dialog": [
    { "type": "recording", "start": "1968-03-12", "duration": 7153,
      "url": "https://...19680312a_416.delivery.mp3" }
  ],
  "analysis": [
    { "type": "wtf_transcription", "dialog": 0, "vendor": "oyez",
      "body": { "segments": [
        { "start": 0.0, "end": 29.0, "speaker": 0, "text": "Number 416, Florence Flast..." },
        { "start": 31.2, "end": 35.2, "speaker": 1, "text": "Mr. Chief Justice and may it please the Court." }
      ] } }
  ],
  "attachments": [
    { "purpose": "case_metadata", "body": { "docket_number": "416", "term": 1967,
      "question": "Did Flast, as a taxpayer, have standing...", "conclusion": "8-to-1 ..." } },
    { "purpose": "lawful_basis", "body": { "lawful_basis": "legitimate_interests" } }
  ]
}

So "search the transcripts" means the analysis bodies, "who argued" means parties, and "what was the case about" means the case_metadata attachment.

Claude Code, one line

terminal
claude mcp add --transport http vcon-scotus https://mcp-scotus.vconic.com/mcp --header "Authorization: Bearer scotus"

Cursor, Windsurf, Cline, any client that sends headers

mcp config JSON
{
  "mcpServers": {
    "vcon-scotus": {
      "type": "http",
      "url": "https://mcp-scotus.vconic.com/mcp",
      "headers": { "Authorization": "Bearer scotus" }
    }
  }
}

claude.ai and ChatGPT connectors (no header support)

Add a custom connector with this URL and no authentication. Same data, read-only.

mcp config JSON
{
  "mcpServers": {
    "vcon-scotus": { "type": "http", "url": "https://mcp-scotus-open.vconic.com/mcp" }
  }
}

Plain REST

shell
curl -s -H "Authorization: Bearer scotus" --get \
  --data-urlencode "q=taxpayer standing" --data "limit=5" \
  https://mcp-scotus.vconic.com/api/v1/vcons/search/content

Full-text search covers the whole transcript. Semantic search covers the subject line only (case name, docket, date), so keyword search is the stronger path here.

Try asking your agent

About the data

Source audio and transcripts are public records of the Supreme Court of the United States, redistributed by Oyez. Each vCon carries a recording dialog pointing at the Court's MP3, a wtf_transcription analysis with one segment per speaker turn, and attachments for case metadata, tags and lawful basis. Corpus, converter and deployment notes: vcon-dev/vcon-supreme-court-arguments. Server: vcon-mcp. A sibling dataset covers IETF working group meetings.

Slide assets: QR (SVG), QR (PNG). The QR encodes this page.