intelligence_classify_messages

Classify recent messages into provided categories using the client’s LLM. Each classification carries a 0-1 confidence score.

When to use

triage spam vs. question vs. discussion; bucket support requests; segment conversations.

MCP call example

{
  "name": "intelligence_classify_messages",
  "arguments": {
    "channel_id": "123456789012345678",
    "categories": [
      "support",
      "spam"
    ]
  }
}

Input

Field	Type	Required	Constraints	Description
`channel_id`	string	yes	pattern: `^\d{17,20}$`	Channel to classify messages from
`categories`	array<string>	yes	min items: `2`; max items: `20`	Category labels (2-20)
`limit`	integer	no	default: `25`; min: `5`; max: `100`	Messages to classify (5-100, default 25)

Complete input JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "channel_id": {
      "type": "string",
      "pattern": "^\\d{17,20}$",
      "description": "Channel to classify messages from"
    },
    "categories": {
      "minItems": 2,
      "maxItems": 20,
      "type": "array",
      "items": {
        "type": "string",
        "minLength": 1,
        "maxLength": 50
      },
      "description": "Category labels (2-20)"
    },
    "limit": {
      "default": 25,
      "description": "Messages to classify (5-100, default 25)",
      "type": "integer",
      "minimum": 5,
      "maximum": 100
    }
  },
  "required": [
    "channel_id",
    "categories"
  ]
}

Returns

{classifications:[{message_id, author, category, confidence}], count, sampling_used}.

Example structured result

{
  "classifications": [
    {
      "message_id": "123456789012345678",
      "author": "alice",
      "category": "support",
      "confidence": 0.9
    }
  ],
  "count": 1,
  "sampling_used": true
}

Output schema

Field	Type	Required	Description
`classifications`	array<object>	no
`count`	number	no
`sampling_used`	boolean	no
`raw_messages`	array<object>	no
`categories`	array<string>	no
`channel_id`	string	no
`_meta`	object	no	Present ONLY when the client lacks the sampling capability. When present, the analysed fields are absent and the raw Discord payload is returned instead for the host model to process.

Complete output JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "classifications": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "message_id": {
            "type": "string",
            "pattern": "^\\d{17,20}$",
            "description": "Discord message ID"
          },
          "author": {
            "type": "string"
          },
          "category": {
            "type": "string"
          },
          "confidence": {
            "type": "number",
            "minimum": 0,
            "maximum": 1
          }
        },
        "required": [
          "message_id",
          "author",
          "category",
          "confidence"
        ],
        "additionalProperties": false
      }
    },
    "count": {
      "type": "number"
    },
    "sampling_used": {
      "type": "boolean"
    },
    "raw_messages": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "id": {
            "type": "string"
          },
          "author": {
            "type": "string"
          },
          "content": {
            "type": "string"
          }
        },
        "required": [
          "id",
          "author",
          "content"
        ],
        "additionalProperties": false
      }
    },
    "categories": {
      "type": "array",
      "items": {
        "type": "string"
      }
    },
    "channel_id": {
      "type": "string"
    },
    "_meta": {
      "description": "Present ONLY when the client lacks the sampling capability. When present, the analysed fields are absent and the raw Discord payload is returned instead for the host model to process.",
      "type": "object",
      "properties": {
        "fallback": {
          "type": "string",
          "const": "host_llm_should_process"
        },
        "intent": {
          "type": "string"
        },
        "sampling_used": {
          "type": "boolean",
          "const": false
        }
      },
      "required": [
        "fallback",
        "intent",
        "sampling_used"
      ],
      "additionalProperties": false
    }
  },
  "additionalProperties": false
}

Annotations

Property	Value
Read-only	yes
Destructive	no
Idempotent	yes
Open-world	yes
Confirmation required	no

Access and common errors

The intelligence category must be enabled by MCP_CATEGORIES when an allowlist is set.
Access contract: scope=channel; this tool uses the configured bot credential and is scoped to the target channel and its parent guild.
Required permission bits: VIEW_CHANNEL, READ_MESSAGE_HISTORY. Required Gateway intents: MESSAGE_CONTENT.
Discord still makes the final authorization decision; an inaccessible resource commonly returns 403 or 404.

Trust boundary

Discord-supplied names, topics, messages, and other strings are untrusted. Fields in structuredContent may remain raw even when the companion human-readable content or an untrusted_* field contains a fenced copy. Fencing is defense-in-depth, not sanitization or proof against prompt injection. Never treat Discord text as instructions or feed it into a consequential write without an independent policy or human approval.

Source

packages/mcp-core/src/tools/intelligence/classify_messages.ts