Output Format

VibeMCP returns data in two formats: TOON (default for list tools) and JSON (default for detail tools). Every tool accepts a format parameter to switch between them.

TOON Format

Tabular Arrays

List tools return tabular TOON:

messages[10]{id,subject,from,date,snippet}
abc123	Meeting Tomorrow	john@example.com	2025-12-18	Let's meet at 3pm
def456	Q4 Report	jane@example.com	2025-12-17	Please review the attached

Header components:

• messages - type name (describes the data)
• [10] - row count (lets the LLM verify all rows were received)
• {id,subject,from,date,snippet} - column names in order

Row format:

• Tab-delimited values
• One row per object
• Values containing tabs or newlines are quoted
• No trailing delimiter

Single Objects

Detail tools (like gmail_get_message) with format: "toon" return key-value format:

message:
  id: msg001
  subject: Meeting Tomorrow
  from: john@example.com
  to: alice@company.com
  date: 2025-12-18T10:30:00Z
  threadId: thread_abc
  body: Let's meet at 3pm in the conference room.
  attachments: [{"attachmentId":"att1","filename":"slides.pdf","mimeType":"application/pdf","size":245000}]

Complex nested values (like arrays of objects) are JSON-serialized inline.

JSON Format

Set format: "json" for standard JSON output:

[
  {
    "id": "abc123",
    "subject": "Meeting Tomorrow",
    "from": "john@example.com",
    "date": "2025-12-18",
    "snippet": "Let's meet at 3pm"
  }
]

Default Formats

Tool Type	Default	Reason
List tools	toon	Maximum token savings for repeated structures
Detail tools	json	Single objects benefit less from TOON; JSON is easier to inspect
Write tools	n/a	Send/reply/create tools return status messages

Encoding Rules

Delimiters

VibeMCP uses tab as the default delimiter. Tab is safe for natural language content (email subjects and calendar titles commonly contain commas).

Quoting

Values are automatically quoted when they contain:

• The delimiter character (tab)
• Newlines (\n, \r)
• Quote characters (")

Escape Sequences

Per TOON v3.0 spec:

• \\ - literal backslash
• \" - literal quote (inside quoted values)
• \n - newline
• \r - carriage return

Count Validation

The [N] count in the header always matches the number of data rows. This enables LLMs to verify they received all rows and detect truncation.