Back to Details

fleet

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

/fleet

/fleet

Fleet orchestration — manage coding agents (Claude Code, Copilot, Amplifier) running across multiple Azure VMs via azlin.
Invoke as 
/fleet <command>
or just describe what you want and Claude will pick the right command.
集群编排——通过azlin管理运行在多台Azure VM上的编码Agent（Claude Code、Copilot、Amplifier）。
可通过
/fleet <command>
调用，或直接描述需求，Claude会自动选择合适的命令。

All Commands

所有命令

Status & Monitoring

状态与监控

CommandWhat it does
fleet status
Quick text overview of all VMs and sessions
fleet scout
Discover, adopt, dry-run reason, show report
fleet snapshot
Point-in-time capture of all managed sessions
fleet watch <vm> <session>
Live snapshot of a single session
fleet observe <vm>
Observe sessions with pattern classification
fleet tui
Interactive Textual TUI dashboard
命令功能
fleet status
所有VM和会话的快速文本概览
fleet scout
发现、接管、试运行原因分析、生成报告
fleet snapshot
捕获所有托管会话的即时状态
fleet watch <vm> <session>
单个会话的实时状态快照
fleet observe <vm>
对会话进行模式分类监控
fleet tui
交互式文本TUI仪表盘

Admiral Control

全局控制

CommandWhat it does
fleet advance
LIVE — reason and execute actions on sessions
fleet dry-run
Show what the admiral would do (no action)
fleet run-once
Single PERCEIVE->REASON->ACT cycle
fleet start
Run autonomous admiral loop
命令功能
fleet advance
实时操作——分析会话并执行对应操作
fleet dry-run
展示全局控制器将执行的操作（无实际执行）
fleet run-once
执行单次PERCEIVE->REASON->ACT循环
fleet start
启动自主全局控制循环

Session Management

会话管理

CommandWhat it does
fleet adopt <vm>
Bring existing tmux sessions under management
fleet auth <vm>
Propagate auth tokens (GitHub, Azure, Claude)
命令功能
fleet adopt <vm>
将现有tmux会话纳入管理
fleet auth <vm>
同步授权令牌（GitHub、Azure、Claude）

Task Management

任务管理

CommandWhat it does
fleet add-task "prompt"
Queue work for the fleet
fleet queue
Show task queue
fleet dashboard
Project-level tracking
fleet report
Generate fleet status report
命令功能
fleet add-task "prompt"
为集群添加待执行任务
fleet queue
查看任务队列
fleet dashboard
项目级跟踪
fleet report
生成集群状态报告

Projects & Knowledge

项目与知识管理

CommandWhat it does
fleet project add/list/remove
Manage fleet projects
fleet project add-issue <proj> <num>
Track a GitHub issue as a project objective
fleet project track-issue <proj>
Sync objectives from GitHub issues by label
fleet graph
Fleet knowledge graph summary
命令功能
fleet project add/list/remove
管理集群项目
fleet project add-issue <proj> <num>
将GitHub议题作为项目目标跟踪
fleet project track-issue <proj>
根据标签从GitHub议题同步目标
fleet graph
集群知识图谱概览

Co-Pilot

副驾驶功能

CommandWhat it does
fleet copilot-status
Show copilot lock/goal state
fleet copilot-log
Show copilot decision history
命令功能
fleet copilot-status
查看Copilot锁定/目标状态
fleet copilot-log
查看Copilot决策历史

Quick Reference

快速参考

User saysCommand
"What are my agents doing?"
fleet scout
"Show me the fleet"
fleet status
"Send next steps to all sessions"
fleet advance
"Advance without confirmation"
fleet advance --force
"Watch what dev/cybergym is doing"
fleet watch dev cybergym
"Add auth to the new VM"
fleet auth <vm>
"Queue this task for the fleet"
fleet add-task "prompt"
"Track issue #42 for myapp"
fleet project add-issue myapp 42
"Sync objectives from GitHub"
fleet project track-issue myapp
"Open the dashboard"
fleet tui
用户提问对应命令
“我的Agent在做什么？”
fleet scout
“展示我的集群状态”
fleet status
“向所有会话发送下一步指令”
fleet advance
“无需确认直接推进”
fleet advance --force
“监控dev/cybergym的操作”
fleet watch dev cybergym
“为新VM添加授权”
fleet auth <vm>
“为集群添加这个任务”
fleet add-task "prompt"
“跟踪myapp的#42号议题”
fleet project add-issue myapp 42
“从GitHub同步目标”
fleet project track-issue myapp
“打开仪表盘”
fleet tui

Key Options

核心选项

fleet scout   [--session vm:session] [--vm VM] [--skip-adopt] [--incremental] [--save PATH]
fleet advance [--session vm:session] [--vm VM] [--force] [--save PATH]
fleet dry-run [--vm VM ...] [--backend auto|anthropic|copilot|litellm]
fleet adopt   <vm> [--sessions s1 s2]
fleet watch   <vm> <session> [--lines 30]
fleet auth    <vm> [--services github azure claude]
fleet add-task "prompt" [--priority high] [--repo URL]
fleet start   [--interval 300] [--max-cycles 10]
fleet scout   [--session vm:session] [--vm VM] [--skip-adopt] [--incremental] [--save PATH]
fleet advance [--session vm:session] [--vm VM] [--force] [--save PATH]
fleet dry-run [--vm VM ...] [--backend auto|anthropic|copilot|litellm]
fleet adopt   <vm> [--sessions s1 s2]
fleet watch   <vm> <session> [--lines 30]
fleet auth    <vm> [--services github azure claude]
fleet add-task "prompt" [--priority high] [--repo URL]
fleet start   [--interval 300] [--max-cycles 10]

Environment

环境变量

VariableRequired for
AZLIN_PATH
All commands (auto-detected if azlin on PATH)
ANTHROPIC_API_KEY
scout, advance, dry-run, run-once, start
变量用途
AZLIN_PATH
所有命令都需要（若azlin在PATH中可自动检测）
ANTHROPIC_API_KEY
scout、advance、dry-run、run-once、start命令需要

Admiral Configuration

全局控制器配置

  • Model: Claude Opus 4.6 (
    claude-opus-4-6
    )
  • Max output tokens: 128,000 (reasoning JSON)
  • Context gathered per session: full tmux scrollback + first 50 & last 200 transcript messages
  • Safety: 57 dangerous-input patterns blocked (with safe allow-list), confidence thresholds (60% send_input, 80% restart), --confirm default on advance
  • Docs: 
    docs/fleet-orchestration/ADMIRAL_REASONING.md
  • 模型: Claude Opus 4.6 (
    claude-opus-4-6
    )
  • 最大输出令牌数: 128,000（推理JSON）
  • 每个会话收集的上下文: 完整tmux回滚记录 + 前50条和后200条对话消息
  • 安全设置: 拦截57种危险输入模式（含安全白名单），置信度阈值（60%发送输入，80%重启），advance命令默认启用--confirm确认
  • 文档: 
    docs/fleet-orchestration/ADMIRAL_REASONING.md

Status Icons

状态图标

IconStatusMeaning
[~]
thinkingAgent is actively processing
[>]
runningAgent producing output
[.]
idleAgent at prompt, waiting for direction
[X]
shell (dead agent)No agent detected in this session
[Z]
suspendedAgent backgrounded but alive
[!]
errorError detected in session output
[+]
completedAgent finished its task
[?]
waiting inputAgent asked a question, awaiting response
图标状态含义
[~]
思考中Agent正在主动处理任务
[>]
运行中Agent正在生成输出
[.]
空闲Agent处于提示符状态，等待指令
[X]
无Agent（会话失效）该会话中未检测到Agent
[Z]
已挂起Agent被后台运行但仍存活
[!]
错误会话输出中检测到错误
[+]
已完成Agent已完成任务
[?]
等待输入Agent提出问题，等待响应

Performance & Architecture

性能与架构

  • Sequential VM polling: Session discovery from 
    azlin list
    (no SSH). Pane capture via sequential SSH
  • Cached SSH output: Scout caches Phase 1 tmux captures for Phase 3 reasoning (no double-poll)
  • Incremental scout: 
    --incremental
    flag skips unchanged sessions using 
    ~/.amplihack/fleet/last_scout.json
  • Bastion tunnel reuse: Reuses existing SSH tunnels via 
    get_existing_tunnels()
    instead of creating new ones
  • PR URL detection: Uses 
    gh pr list
    on remote VM for reliable PR detection from git state
  • Health metrics in reasoning: 
    fleet_health.py
    wired into 
    SessionContext
    for admiral decisions
  • Unified status classifier: Single canonical classifier in 
    _status.py
    (no dual TUI/CLI divergence)
  • Modular CLI: Commands split across 
    _cli_session_ops.py
    , 
    _cli_scout_advance.py
    , 
    _cli_formatters.py
    (each under 400 LOC)
  • Project grouping: Scout report groups sessions by registered project with open objectives displayed
  • 顺序VM轮询: 通过
    azlin list
    发现会话（无需SSH），通过顺序SSH捕获面板内容
  • SSH输出缓存: Scout在第1阶段缓存tmux捕获结果用于第3阶段推理（避免重复轮询）
  • 增量探测: 
    --incremental
    标志会跳过未更改的会话，使用
    ~/.amplihack/fleet/last_scout.json
    记录
  • 堡垒机隧道复用: 通过
    get_existing_tunnels()
    复用现有SSH隧道，而非创建新隧道
  • PR URL检测: 在远程VM上使用
    gh pr list
    从git状态中可靠检测PR
  • 推理中的健康指标: 
    fleet_health.py
    接入
    SessionContext
    用于全局控制器决策
  • 统一状态分类器: 
    _status.py
    中包含单一标准分类器（TUI与CLI无分歧）
  • 模块化CLI: 命令拆分到
    _cli_session_ops.py
    、
    _cli_scout_advance.py
    、
    _cli_formatters.py
    （每个文件不超过400行代码）
  • 项目分组: Scout报告按已注册项目将会话分组，并显示未完成目标

How to Run

运行方式

Execute via Bash:
bash
fleet <command> [options]
通过Bash执行：
bash
fleet <command> [options]

Presenting Results

结果展示

After running 
fleet scout
or 
fleet advance
, present the output to the user as:
  1. Summary table — reformat the CLI output into a clean markdown table with columns: VM | Session | Status | Action | Conf | Summary
  2. Proposed inputs — for any 
    send_input
    decisions, show what the admiral wants to send
  3. Follow-up commands — always end with the actionable next steps from the report:
    • fleet advance
      to send next command to all sessions
    • fleet advance --confirm
      to review each before executing
    • fleet advance --session vm:session
      to advance one specific session
    • fleet watch <vm> <session>
      to inspect specific sessions
运行
fleet scout
或
fleet advance
后，按以下格式向用户展示输出：
  1. 摘要表格——将CLI输出重新格式化为清晰的Markdown表格，包含列： VM | 会话 | Status | 操作 | 置信度 | 摘要
  2. 建议输入——对于所有
    send_input
    决策，展示全局控制器想要发送的内容
  3. 后续命令——始终在报告末尾提供可执行的下一步操作：
    • fleet advance
      向所有会话发送下一条命令
    • fleet advance --confirm
      先逐个审核再执行
    • fleet advance --session vm:session
      推进单个指定会话
    • fleet watch <vm> <session>
      查看指定会话详情