mcp_pipeline

Meta Writes to Discord

Execute a sequence of MCP tool calls in one request. Variables from earlier steps interpolate into later steps via {{step_id.path}}.

When to use

when a workflow needs ≥2 sequential calls (e.g., list channels → find by name → send message). Reduces N round-trips to 1.

When not to use

parallel-safe independent calls - issue them as separate tools/call requests. Long-running batch ops - use the dedicated bulk tool (Plan 7+) so each operation can fail independently.

Step shape

{id?, tool, args, save_as?, if?, continue_on_error?}. args may contain {{step_id.path}} placeholders. if is a path check; the step skips when the path resolves to falsy.

Limits

max 20 steps per pipeline. Nested mcp_pipeline rejected (no recursion).

MCP call example

Tool-authored example

{steps:[
  {id:"channels", tool:"channels_list", args:{guild_id:"123456789012345678"}},
  {id:"send",     tool:"messages_send", args:{channel_id:"{{channels.channels[0].id}}", content:"hi"}}
]}

{
  "name": "mcp_pipeline",
  "arguments": {
    "steps": [
      {
        "id": "me",
        "tool": "users_get_current",
        "args": {}
      }
    ]
  }
}

Input

Field	Type	Required	Constraints	Description
steps	array<object>	yes	min items: 1; max items: 20	Ordered steps. Max 20 per pipeline.

Complete input JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "steps": {
      "minItems": 1,
      "maxItems": 20,
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "id": {
            "description": "Reference name for interpolation. Defaults to step_<index>.",
            "type": "string",
            "pattern": "^[a-z][a-z0-9_]{0,63}$"
          },
          "tool": {
            "type": "string",
            "minLength": 1,
            "description": "Tool name to call. Must exist; not mcp_pipeline (no recursion)."
          },
          "args": {
            "type": "object",
            "propertyNames": {
              "type": "string"
            },
            "additionalProperties": {},
            "description": "Tool arguments. String fields support {{step_id.path}} interpolation."
          },
          "save_as": {
            "description": "Save result under this variable name in addition to step ID.",
            "type": "string",
            "minLength": 1
          },
          "if": {
            "description": "Path to truthy check. Step skipped if path resolves falsy. Example: \"{{step1.count}}\".",
            "type": "string",
            "minLength": 1
          },
          "continue_on_error": {
            "default": false,
            "description": "If true, pipeline continues after this step fails. Default aborts.",
            "type": "boolean"
          }
        },
        "required": [
          "tool",
          "args"
        ]
      },
      "description": "Ordered steps. Max 20 per pipeline."
    }
  },
  "required": [
    "steps"
  ]
}

Returns

{steps:[{id, tool, status, result?, error?, duration_ms}], variables, total_duration_ms, aborted}. Each step status is success | error | skipped.

Example structured result

{
  "steps": [
    {
      "id": "123456789012345678",
      "tool": "example",
      "status": "success",
      "duration_ms": 1
    }
  ],
  "variables": {},
  "total_duration_ms": 1,
  "aborted": true
}

Output schema

Field	Type	Required	Constraints	Description
steps	array<object>	yes
variables	object	yes
total_duration_ms	number	yes
aborted	boolean	yes

Complete output JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "steps": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "id": {
            "type": "string"
          },
          "tool": {
            "type": "string"
          },
          "status": {
            "type": "string",
            "enum": [
              "success",
              "error",
              "skipped"
            ]
          },
          "result": {},
          "error": {
            "type": "object",
            "properties": {
              "code": {
                "type": "string"
              },
              "message": {
                "type": "string"
              },
              "retriable": {
                "type": "boolean"
              }
            },
            "required": [
              "code",
              "message",
              "retriable"
            ],
            "additionalProperties": false
          },
          "duration_ms": {
            "type": "number"
          }
        },
        "required": [
          "id",
          "tool",
          "status",
          "duration_ms"
        ],
        "additionalProperties": false
      }
    },
    "variables": {
      "type": "object",
      "propertyNames": {
        "type": "string"
      },
      "additionalProperties": {}
    },
    "total_duration_ms": {
      "type": "number"
    },
    "aborted": {
      "type": "boolean"
    }
  },
  "required": [
    "steps",
    "variables",
    "total_duration_ms",
    "aborted"
  ],
  "additionalProperties": false
}

Annotations

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

Access and common errors

• The meta category remains available even when MCP_CATEGORIES is restricted.
• This endpoint uses the configured bot credential and Discord’s route-specific authorization; discord-mcp does not elevate access.
• An invalid credential returns a 401-class tool error. Insufficient endpoint permission or scope returns 403; inaccessible resources commonly return 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/meta/pipeline.ts