LoopFlowHQ MCP Capability Docs

MCP tool capability list and usage guidance for chat/coding agents.

Prompt contract: /llm/latest · REST docs: /llm/api

Discovery Pattern

1) List available tools from MCP. 2) Read each tool schema/description. 3) Execute supported tools only.

If REST is available, you can also discover capabilities via GET /api/chatgpt-actions/schema.

Registered MCP Tools

  • list_projects
  • list_my_tickets
  • get_ticket
  • get_ticket_messages
  • get_ticket_context
  • next_best_child_story
  • next_best_action
  • create_ticket
  • update_ticket
  • update_ticket_status
  • get_ticket_subtasks
  • update_ticket_subtasks
  • post_ticket_message
  • post_ticket_agent_questions

Required Inputs (common)

Most ticket-scoped tools require ticket as UUID or ticket key (for example, LOOPF-73).

To persist recap quality: use agent_memory, acceptance_criteria, and structured chat posts.

To persist PR URL: set pull_request_url via update_ticket.

Examples

update_ticket({
  ticket: "LOOPF-73",
  acceptance_criteria: "1. ...\n2. ...",
  agent_memory: "Session recap...",
  phase: "refine",
  phase_state: "in_progress"
})
update_ticket({
  ticket: "LOOPF-73",
  pull_request_url: "https://github.com/org/repo/pull/123"
})
post_ticket_agent_questions({
  ticket: "LOOPF-73",
  questions: [
    {
      id: "risk_area",
      question: "What is the highest-risk assumption?",
      proposed_answers: ["Adoption", "Latency", "Data quality"]
    }
  ]
})