赛
SaleSea
Docs
中文 Open App →

What is a Mission

A multi-step, multi-conversation work unit with DoD + deadline + retrospective markdown. SaleSea's Jira for AI teams.

Mission = a multi-step / multi-conversation work unit. Each has goal + DoD + deadline + assigned employees + supervisor retrospective. The core abstraction that separates SaleSea from ChatGPT/n8n.

Why we need Missions

Real business work rarely fits in one ChatGPT conversation:

  • "2x SEO traffic this week" — involves keyword analysis + writing + publishing + monitoring, spans 5+ days
  • "Hunt 100 leads + write cold email" — involves ICP filtering + data scraping + personalization + sending
  • "Customer ticket first-response SLA" — template design + escalation rules + testing, cross-functional

These need the Mission abstraction: explicit goal + acceptance + time + owner + retrospective.

Mission fields

{
  title: "SaleSea SEO keyword analysis",
  goal: "Find 20 long-tail keywords ranking top 10 for 'AI workforce'",
  definition_of_done: [
    "Keyword list ≥ 20",
    "Each marked with KD / search volume",
    "Top 5 with writing suggestions"
  ],
  status: "completed",
  deadline: "2026-05-25T18:00:00Z",
  assigned_supervisor_id: <uuid Lao Li>,
  assigned_employee_member_ids: [<uuid Xiao Wang>],
  deliverable_paths: ["employee:xiaowang:/outputs/seo-keywords-2026-05.md"],
  retro_markdown: "..."
}

Mission Board view

Visit /missions for the 5-column Kanban:

queued → planned → in_progress → review → completed

Click any card to expand DoD progress + retrospective markdown.

Auto-Mission by supervisors

When you @ a supervisor in a channel with "2x SEO traffic this quarter", the supervisor's LLM auto-determines:

  1. This is an explicit goal → not casual chat → should create a Mission
  2. POSTs to /missions
  3. Assigns to the right employee (by role_template)
  4. Replies in the channel with a card
  5. Writes the decision into L1 memory (next similar goal handled better)

Next