generate_post_text

Generate post text

Generate scrubbed LinkedIn post text from an idea with usage tracking.

Authorization and permissions

Requires a valid delegated OAuth token and access to the selected workspace.

OAuth scope: mcp:ai
Workspace requirement: Authorized workspace and current membership required

Current member; existing AI usage quota applies.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "idea": {
      "description": "Idea.",
      "maxLength": 10000,
      "minLength": 1,
      "type": "string"
    },
    "team_id": {
      "description": "Consented workspace ID.",
      "minimum": 1,
      "type": "integer"
    }
  },
  "required": [
    "team_id",
    "idea"
  ],
  "type": "object"
}

Result schema

{
  "additionalProperties": false,
  "properties": {
    "content": {
      "description": "Content.",
      "type": "string"
    }
  },
  "required": [
    "content"
  ],
  "type": "object"
}

Example arguments

{
  "idea": "What we learned from interviewing ten new customers about onboarding",
  "team_id": 42
}

Example result

{
  "content": "Ten customer interviews changed how we think about onboarding.\n\nThe biggest lesson: ask about the last real attempt, not an ideal future."
}

Errors and retry

The connection is limited to 120 tool calls per user per minute and AI tools to 30 calls per user per hour. Invalid or cross-workspace identifiers return not found; revoked membership/scope returns permission denied; invalid input and stale versions return validation errors. Existing product limits still apply; there is no separate MCP subscription quota.

Request metadata (actor, connection, workspace, tool, outcome, duration, and correlation ID) is logged; arguments and results are not. Retry reads freely. Retry a mutation only when marked idempotent; otherwise read current state before deciding whether to repeat it.

Effects, limits, and source

Workspace roles, subscription and AI usage limits, approval rules, and server rate limits are enforced at call time. Rate-limit responses identify when to retry; scheduled work returns state rather than holding the request open.

ai.generation.generate_from_idea