claude-emporium

MCP开发工具

提供方 Vvkmnn · 包名 claude-emporium

开发工具 stdio · 本地进程 未标注 社区 / 第三方 Claude DesktopCursorClineClaude Code
查看接入方式

🏛 [UNDER CONSTRUCTION] A (roman) claude plugin marketplace

一句话选型结论

按资料字段整理

claude-emporium 属于开发工具类 MCP 服务器,stdio · 本地进程,鉴权方式未标注,可直接接入 Claude Desktop、Cursor、Cline 等,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类开发工具 传输方式stdio · 本地进程 鉴权未标注 部署方式本地部署
GitHub 星标 113 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

开发工具
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

claude-emporium ai ai-agents anthropic automation

支持客户端

适用场景

claude-emporiumaiai-agents

安装与配置

stdio · 本地进程 未标注
客户端在本机拉起子进程,数据不出本机
npx -y claude-emporium

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Cross-platform,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
代码与仓库操作

在对话里直接查代码、提 issue、走 PR,省掉来回切工具。

本地命令与脚本

把构建、测试、格式化这类重复命令交给模型执行。

接口联调与测试

让模型调你的接口、跑一遍用例,快速定位问题。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Cursor、Cline,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在开发工具场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Cursor 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
claude-emporium当前 未标注 stdio · 本地进程 本地部署 有 2026-04-08
Git 838e2ec1 12 stdio · 本地进程 本地部署 有 2026-05-17
Markitdown MCP 2c653a5f 1 stdio · 本地进程 本地部署 有 2026-05-17
JavaGuide 未标注 stdio · 本地进程 本地部署 有 2026-04-11
awesome-claude-skills 未标注 stdio · 本地进程 本地部署 有 2026-04-26
Context7 831acf40 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

开发工具
字段来自项目仓库与公开索引,中文说明由本站标注
提供方Vvkmnn上游标注的维护方 Vvkmnn
分类开发工具本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式未标注远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时claude-emporium · Cross-platform本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-04-08上游最后核验时间

详细介绍

30.9 KB
来自项目仓库与上游目录的原文,未做删改

claude emporium

a (roman) plugin marketplace for claude code

install plugins synergy architecture contributing

roma victrix


Six plugins that wrap standalone MCP servers with automation hooks, commands, and skills. No code duplication: plugins tell claude when to act, MCPs handle how.```
╔═╤═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╤═╗
║ │ │ ║
║ │ PLUGINS │ ║
║ │ │ ║
║ │ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ │ ║
║ │ │ PRAETORIAN │ │ HISTORIAN │ │ ORACLE │ │ GLADIATOR │ │ VIGIL │ │ ORATOR │ │ ║
║ │ │ context guard │ │ session memory │ │ tool discovery │ │ learn & adapt │ │ file recovery │ │ prompt rhetoric │ │ ║
║ │ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ │ ║
║ │ │ hooks │ │ hooks │ │ hooks │ │ hooks │ │ hooks │ │ hooks │ │ ║
║ │ │ · pre-plan │ │ · pre-websearch │ │ · pre-plan │ │ · post-error │ │ · pre-bash │ │ · pre-task │ │ ║
║ │ │ · pre-compact │ │ · pre-plan │ │ · post-error │ │ · stop │ │ │ │ │ │ ║
║ │ │ · post-research │ │ · pre-task │ │ │ │ │ │ commands │ │ commands │ │ ║
║ │ │ · subagent-stop │ │ · post-error │ │ commands │ │ commands │ │ · /save-vigil │ │ · /reprompt- │ │ ║
║ │ │ │ │ │ │ · /search-oracle │ │ · /review- │ │ · /restore-vigil │ │ orator │ │ ║
║ │ │ commands │ │ commands │ │ │ │ gladiator │ │ │ │ │ │ ║
║ │ │ · /compact- │ │ · /search- │ │ │ │ │ │ │ │ │ │ ║
║ │ │ praetorian │ │ historian │ │ │ │ │ │ │ │ │ │ ║
║ │ │ · /restore- │ │ │ │ │ │ │ │ │ │ │ │ ║
║ │ │ praetorian │ │ │ │ │ │ │ │ │ │ │ │ ║
║ │ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ │ ║
║ │ │ │ │ │ │ │ │ ║
║ │ ▼ ▼ ▼ ▼ ▼ ▼ │ ║
║ │ │ ║
║ │ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ │ ║
║ │ │ praetorian-mcp │ │ historian-mcp │ │ oracle-mcp │ │ gladiator-mcp │ │ vigil-mcp │ │ orator-mcp │ │ ║
║ │ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ │ ║
║ │ │ save_context │ │ search_convos │ │ search │ │ observe │ │ vigil_save │ │ orator_optimize │ │ ║
║ │ │ · snapshot before │ │ · full-text across │ │ · query 17 sources │ │ · record patterns │ │ · named checkpoint │ │ · score 7 dims │ │ ║
║ │ │ compaction │ │ all sessions │ │ in parallel │ │ │ │ │ │ · apply 8 techs │ │ ║
║ │ │ │ │ │ │ │ │ reflect │ │ vigil_list │ │ · restructure │ │ ║
║ │ │ restore_context │ │ get_error_solns │ │ browse │ │ · cluster and │ │ · show checkpoints │ │ │ │ ║
║ │ │ · load previous │ │ · how errors were │ │ · by category or │ │ recommend │ │ │ │ ── ── ── ── ── │ │ ║
║ │ │ session state │ │ resolved │ │ popularity │ │ │ │ vigil_diff │ │ dimensions: │ │ ║
║ │ │ │ │ │ │ │ │ ── ── ── ── ── │ │ · preview changes │ │ clarity │ │ ║
║ │ │ search_compactns │ │ find_similar │ │ sources │ │ storage: │ │ │ │ specificity │ │ ║
║ │ │ · find past saves │ │ · related past │ │ · list registries │ │ .claude/ │ │ vigil_restore │ │ structure │ │ ║
║ │ │ │ │ questions │ │ and status │ │ gladiator/ │ │ · restore files │ │ context │ │ ║
║ │ │ list_compactions │ │ │ │ │ │ │ │ │ │ examples │ │ ║
║ │ │ · browse recent │ │ find_file_context │ │ ── ── ── ── ── │ │ │ │ vigil_delete │ │ constraints │ │ ║
║ │ │ snapshots │ │ · track changes │ │ smithery · glama │ │ │ │ · remove checkpoint│ │ tone │ │ ║
║ │ │ │ │ │ │ npm · github │ │ │ │ │ │ │ │ ║
║ │ │ ── ── ── ── ── │ │ find_tool_pattns │ │ awesome-mcp │ │ │ │ ── ── ── ── ── │ │ ── ── ── ── ── │ │ ║
║ │ │ storage: │ │ · agent workflows │ │ mcp-registry │ │ │ │ storage: │ │ in-memory │ │ ║
║ │ │ .claude/ │ │ │ │ + 11 more │ │ │ │ .claude/ │ │ zero storage │ │ ║
║ │ │ praetorian/ │ │ search_plans │ │ │ │ │ │ vigil/ │ │ │ │ ║
║ │ │ │ │ · past plans │ │ in-memory cache │ │ │ │ │ │ │ │ ║
║ │ │ │ │ │ │ zero storage │ │ │ │ │ │ │ │ ║
║ │ │ │ │ list_recent │ │ │ │ │ │ │ │ │ │ ║
║ │ │ │ │ · recent sessions │ │ │ │ │ │ │ │ │ │ ║
║ │ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ │ ║
║ │ │ ║
║ │ MCP SERVERS │ ║
╚═╧═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╧═╝


## install

Each plugin self-configures on install. MCP servers, hooks, commands, skills, and sibling detection all register automatically. All hooks default to enabled. See configuration to disable specific hooks or tune behavior.

**Requirements:**

**Marketplace:**```bash
# Add this marketplace
/plugin marketplace add Vvkmnn/claude-emporium

Plugins:```bash

Each registers its MCP server, skills, hooks, settings, and detects siblings automatically

/plugin install claude-praetorian@claude-emporium
/plugin install claude-historian@claude-emporium
/plugin install claude-oracle@claude-emporium
/plugin install claude-gladiator@claude-emporium
/plugin install claude-vigil@claude-emporium
/plugin install claude-orator@claude-emporium


**MCP servers:**```bash
# NOT REQUIRED: Each plugin will install their MCP server by themself
# However, you can choose to install just the MCP servers if you like
# Standalone has no hooks or synergy, but each ships with an optionally installable skill
# claude mcp add praetorian -- npx claude-praetorian-mcp  # https://github.com/Vvkmnn/claude-praetorian-mcp
# claude mcp add historian -- npx claude-historian-mcp    # https://github.com/Vvkmnn/claude-historian-mcp
# claude mcp add oracle -- npx claude-oracle-mcp          # https://github.com/Vvkmnn/claude-oracle-mcp
# claude mcp add gladiator -- npx claude-gladiator-mcp    # https://github.com/Vvkmnn/claude-gladiator-mcp
# claude mcp add vigil -- npx claude-vigil-mcp            # https://github.com/Vvkmnn/claude-vigil-mcp
# claude mcp add orator -- npx claude-orator-mcp          # https://github.com/Vvkmnn/claude-orator-mcp

plugins

praetorian

Context Guard. Saves and restores valuable context before compaction, after research, and when subagents complete.

mcp: claude-praetorian-mcp (storage at .claude/praetorian/)

commands:

  • /compact-praetorian · save insights after research, decisions, or subagent work
  • /restore-praetorian · load relevant past compactions at session start

tools:

  • praetorian_compact · save structured context: research, decisions, flow analysis
  • praetorian_restore · search or list recent compactions by query or type

hooks:

  • PreToolUse on EnterPlanMode · lists prior compactions before planning
  • PreCompact on * · saves context before compaction resets it
  • PostToolUse on WebFetch / WebSearch · prompts to compact research findings
  • SubagentStop on * · prompts to compact subagent results

historian

Session searcher. Checks past sessions before you do redundant research, planning, or debugging to find solutions or insights we can recycle.

mcp: claude-historian-mcp (9 search tools across conversations, errors, files, plans)

commands:

  • /search-historian · search past sessions for solutions, patterns, or context

tools:

  • search_conversations · search conversation history, .claude files, and tasks
  • find_file_context · find all conversations related to a specific file
  • find_similar_queries · find previous similar questions with enhanced matching
  • get_error_solutions · find solutions for specific errors
  • list_recent_sessions · browse recent sessions with activity summaries
  • extract_compact_summary · get intelligent summary of a session with key insights
  • find_tool_patterns · analyze tool usage patterns and workflows
  • search_plans · search plan files for past approaches and decisions
  • search_config · search .claude config files for guidance and patterns

hooks:

  • PreToolUse on WebSearch / WebFetch · checks history before web research
  • PreToolUse on EnterPlanMode · searches past plans before planning
  • PreToolUse on Task · checks tool patterns before launching agents
  • PostToolUse on Bash · suggests error solutions after failures

oracle

Tool recommender. Searches 17 sources in parallel to find relevant skills, plugins, and MCP servers.

mcp: claude-oracle-mcp (smithery, glama, npm, github, awesome-mcp-servers, and more)

commands:

  • /search-oracle · search all sources for skills, plugins, or MCP servers

tools:

  • search · search by query with optional type and semantic filters
  • browse · browse by category or popularity
  • sources · show available data sources and their status

hooks:

  • PreToolUse on EnterPlanMode · searches for relevant tools before planning
  • PostToolUse on Bash · searches for tools that solve errors

gladiator

Continuous learner. Observes tool failures and prompts reflection at session end to evolve rules, hooks, and skills incrementally.

mcp: claude-gladiator-mcp (storage at .claude/gladiator/)

commands:

  • /review-gladiator · review and cluster observations into actionable patterns

tools:

  • gladiator_observe · record a pattern worth learning from, deduplicates by hash
  • gladiator_reflect · query and cluster observations with recommendations

hooks:

  • PostToolUse on Bash / Edit / Write · observes failure patterns (silent on success)
  • Stop on * · prompts reflection if unprocessed observations exist

vigil

Quick saver. Saves checkpoints before dangerous operations, diffs changes, restores files safely to augment /rewind.

mcp: claude-vigil-mcp (storage at .claude/vigil/)

commands:

  • /save-vigil · create a named checkpoint of tracked files
  • /restore-vigil · restore files from a previous checkpoint

tools:

  • vigil_save · save a named checkpoint with optional description
  • vigil_list · list checkpoints, drill into one for file details
  • vigil_diff · diff current files against a checkpoint
  • vigil_restore · restore files from a checkpoint
  • vigil_delete · delete a checkpoint or all checkpoints

hooks:

  • PreToolUse on Bash · auto-quicksaves before destructive commands (rm, mv, etc.)

orator

Prompt coach. Deterministically scores and restructures prompts using Anthropic best practices. No LLM calls, no network, in-memory only.

mcp: claude-orator-mcp (in-memory, zero storage)

commands:

  • /reprompt-orator · score and optimize a prompt with detailed feedback

tools:

  • orator_optimize · analyze and optimize a prompt, returns score metrics and applied techniques

hooks:

  • PreToolUse on Task · suggests optimization for under-specified subagent prompts

architecture

Each plugin is a thin wrapper. hooks inject prompts that trigger MCP tools at high-impact moments. plugins contain no business logic: they tell claude when to search, save, or restore. the MCP servers handle the actual work.```
claude-emporium/
├── .claude-plugin/
│ ├── marketplace.json plugin registry (6 plugins)
│ └── plugin.json root manifest
│
├── plugins/
│ ├── claude-praetorian/ context guard → praetorian-mcp
│ │ ├── .claude-plugin/plugin.json declares praetorian MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json hook event bindings
│ │ │ ├── post-research.js compact after WebSearch/WebFetch
│ │ │ ├── post-subagent.js compact after Task completes
│ │ │ ├── pre-compact.js remind to compact before context reset
│ │ │ └── pre-plan.js list prior compactions before planning
│ │ ├── commands/
│ │ │ ├── compact-praetorian.md /compact
│ │ │ └── restore-praetorian.md /compact-status
│ │ ├── skills/claude-praetorian/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-historian/ session memory → historian-mcp
│ │ ├── .claude-plugin/plugin.json declares historian MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ ├── pre-websearch.js check history before web research
│ │ │ ├── pre-planning.js search past plans before planning
│ │ │ ├── pre-task.js check tool patterns before agents
│ │ │ └── post-error.js suggest past error solutions
│ │ ├── commands/search-historian.md /history
│ │ ├── skills/claude-historian/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-oracle/ tool discovery → oracle-mcp
│ │ ├── .claude-plugin/plugin.json declares oracle MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ ├── pre-planning.js search registries before planning
│ │ │ └── post-error.js find tools that solve errors
│ │ ├── commands/search-oracle.md /discover
│ │ ├── skills/claude-oracle/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-gladiator/ learn & adapt → gladiator-mcp
│ │ ├── .claude-plugin/plugin.json declares gladiator MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ ├── stop.js prompt reflection at session end
│ │ │ └── post-error.js observe failure patterns
│ │ ├── commands/review-gladiator.md /observe · /reflect
│ │ ├── skills/claude-gladiator/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-vigil/ file recovery → vigil-mcp
│ │ ├── .claude-plugin/plugin.json declares vigil MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ └── pre-bash.js quicksave before rm/reset/checkout
│ │ ├── commands/
│ │ │ ├── save-vigil.md /save
│ │ │ └── restore-vigil.md /restore · /snapshots
│ │ ├── skills/claude-vigil/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-orator/ prompt rhetoric → orator-mcp
│ │ ├── .claude-plugin/plugin.json declares orator MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ └── pre-task.js optimize prompts before subagents
│ │ ├── commands/reprompt-orator.md /optimize
│ │ ├── skills/claude-orator/SKILL.md
│ │ └── README.md
│ │
│ └── shared/
│ └── utils.js loadSettings · siblings · readStdin · emit
│
├── mcp/
│ └── .mcp.json all 6 MCP servers (standalone install)
│
├── logo/ plugin SVG icons
├── .github/ issue templates · PR template · CI workflows
├── scripts/readme.py README generation
└── README.md

each plugin.json declares its MCP server:
{ "mcpServers": { "praetorian": { "command": "npx", "args": ["claude-praetorian-mcp"] } } }

each hook script follows the same pattern:
readStdin() → loadSettings() → siblings() → emit(<system-reminder>)

runtime storage:
project/.claude/praetorian/ context snapshots
project/.claude/gladiator/ observations + reflections
project/.claude/vigil/ file checkpoints
~/.claude/settings.json enabledPlugins (auto) · claude-emporium (manual)


**zero overhead, by design:**

- no external API calls. no network requests from hooks or plugins
- no background processes. hooks run inline and exit (~1ms each)
- no databases. praetorian, gladiator, and vigil store to project-local dirs (`.claude/`). oracle, historian, and orator compute in-memory with TTL caching
- no dependencies beyond Node.js (ships with Claude Code)
- all hook scripts are pure Node.js, single-purpose, and stateless
- built for claude, by claude

## configuration

**Sibling detection:** `/plugin install` adds each plugin to `enabledPlugins` in `~/.claude/settings.json`. Hooks read this to detect siblings and coordinate automatically. For example, historian&#x27;s pre-research hook will note &quot;praetorian will compact this&quot; when both are installed, and oracle&#x27;s error hook will mention &quot;historian can search past solutions&quot; if historian is present. This is what sibling detection looks like:```jsonc
// ~/.claude/settings.json
{
  // ...model, permissions, apiKey, etc.
  &quot;enabledPlugins&quot;: {
    // ...other installed plugins
    &quot;claude-praetorian@claude-emporium&quot;: true, // detected as sibling by all other plugins
    &quot;claude-historian@claude-emporium&quot;: true, // detected as sibling by all other plugins
    &quot;claude-oracle@claude-emporium&quot;: true,
    // ...added automatically by /plugin install
  },
  // ...
}

Hook settings: every hook checks a boolean setting before running. All default to true. To disable any hook, add overrides under the claude-emporium key in ~/.claude/settings.json:```jsonc
// ~/.claude/settings.json
{
"claude-emporium": {
"suggest_siblings": false, // stop suggesting uninstalled emporium plugins

&quot;claude-praetorian&quot;: {
  &quot;auto_compact_research&quot;: false, // don&#x27;t prompt to compact after WebFetch/WebSearch
  &quot;auto_compact_subagent&quot;: false, // don&#x27;t prompt to compact after subagent completes
  &quot;check_compactions_before_plan&quot;: false, // don&#x27;t list prior compactions before planning
  &quot;remind_compact&quot;: false, // don&#x27;t remind to compact before context reset
},
&quot;claude-historian&quot;: {
  &quot;search_before_web&quot;: false, // don&#x27;t check history before web research
  &quot;search_before_plan&quot;: false, // don&#x27;t search past plans before planning
  &quot;search_before_task&quot;: false, // don&#x27;t check tool patterns before launching agents
  &quot;search_after_error&quot;: false, // don&#x27;t suggest error solutions after failures
},
&quot;claude-oracle&quot;: {
  &quot;search_before_plan&quot;: false, // don&#x27;t search for relevant tools before planning
  &quot;search_after_error&quot;: false, // don&#x27;t search for tools that solve errors
},
&quot;claude-gladiator&quot;: {
  &quot;observe_after_failure&quot;: false, // don&#x27;t record failure patterns
  &quot;reflect_before_stop&quot;: false, // don&#x27;t prompt reflection at session end
},
&quot;claude-vigil&quot;: {
  &quot;auto_quicksave&quot;: false, // don&#x27;t quicksave before destructive commands
},
&quot;claude-orator&quot;: {
  &quot;optimize_subagent_prompts&quot;: false, // don&#x27;t suggest prompt optimization before Task
},

},
}


Only include the settings you want to change. Omitted keys keep their defaults (`true`). The full defaults are defined in `shared/utils.js`.

## synergy

Each plugin works standalone. when multiple are installed, they **detect siblings at runtime and coordinate**, usually without any extra configuration.```mermaid
%%{init: {&#x27;theme&#x27;: &#x27;base&#x27;, &#x27;htmlLabels&#x27;: true, &#x27;themeVariables&#x27;: {&#x27;primaryColor&#x27;: &#x27;#f5f5f5&#x27;, &#x27;primaryTextColor&#x27;: &#x27;#1a1a1a&#x27;, &#x27;primaryBorderColor&#x27;: &#x27;#cccccc&#x27;, &#x27;lineColor&#x27;: &#x27;#555555&#x27;, &#x27;edgeLabelBackground&#x27;: &#x27;#f0f0f0&#x27;, &#x27;fontSize&#x27;: &#x27;16px&#x27;}, &#x27;flowchart&#x27;: {&#x27;nodeSpacing&#x27;: 80, &#x27;rankSpacing&#x27;: 70, &#x27;curve&#x27;: &#x27;basis&#x27;, &#x27;padding&#x27;: 20}}}%%
flowchart TD
    P[&quot;⚜️&lt;br/&gt;PRAETORIAN&lt;br/&gt;context guard&quot;]:::gold
    H[&quot;📜&lt;br/&gt;HISTORIAN&lt;br/&gt;session memory&quot;]:::scroll
    V[&quot;🏺&lt;br/&gt;VIGIL&lt;br/&gt;file recovery&quot;]:::amphora
    O[&quot;🔮&lt;br/&gt;ORACLE&lt;br/&gt;tool discovery&quot;]:::crystal
    G[&quot;⚔️&lt;br/&gt;GLADIATOR&lt;br/&gt;learn &amp; adapt&quot;]:::steel
    R[&quot;🪶&lt;br/&gt;ORATOR&lt;br/&gt;prompt rhetoric&quot;]:::quill

    P ----|&quot;  plans &amp; context  &quot;| H
    V ----|&quot;  quicksave  &quot;| P
    P ----|&quot;  compactions  &quot;| O
    H ----|&quot;  solutions + tools  &quot;| O
    O ----|&quot;  error tools  &quot;| G
    H ----|&quot;  past solutions  &quot;| G
    R ----|&quot;  scored prompts  &quot;| H
    R ----|&quot;  compact results  &quot;| P
    R ----|&quot;  find tools  &quot;| O
    R ----|&quot;  observe techniques  &quot;| G
    R ----|&quot;  checkpoint  &quot;| V

    classDef gold fill:#fff8e1,stroke:#f9a825,stroke-width:2px,color:#5d4037
    classDef scroll fill:#fff3e0,stroke:#ef6c00,stroke-width:2px,color:#4e342e
    classDef amphora fill:#fbe9e7,stroke:#e64a19,stroke-width:2px,color:#bf360c
    classDef crystal fill:#f3e5f5,stroke:#8e24aa,stroke-width:2px,color:#4a148c
    classDef steel fill:#eceff1,stroke:#546e7a,stroke-width:2px,color:#263238
    classDef quill fill:#efebe9,stroke:#795548,stroke-width:2px,color:#3e2723

Enhanced behaviors when siblings are detected:

event plugin alone with siblings
plan ⚜️ praetorian lists prior compactions + 📜 past plans, 🔮 finds tools
plan 📜 historian searches past plans + 🔮 also searches for tools
plan 🔮 oracle searches 17 registries + 📜 past plans, ⚜️ compactions
compact ⚜️ praetorian snapshots context + 🔮 discoveries included
error 📜 historian searches past solutions + 🔮 finds tools for this error class
error 🔮 oracle searches for error-solving tools + 📜 checks past solutions
error ⚔️ gladiator observes failure pattern + 📜 enriches with past solutions
research 📜 historian checks history first + ⚜️ compacts findings after
subagent ⚜️ praetorian prompts to compact results + ⚔️ observes subagent outcome patterns
task 📜 historian checks tool patterns + 🔮 suggests tools for the task
task 🪶 orator optimizes vague prompts + 📜 searches past scored prompts
stop ⚔️ gladiator prompts reflection + 📜 and 🔮 deepen analysis
bash 🏺 vigil quicksaves before destructive + ⚜️ preserves context alongside

Detection is automatic via one fs.readFileSync call (~1ms), and falls back gracefully to solo mode if settings are missing or siblings are not dedicated.

contributing

Bug fixes, improvements, and documentation PRs are welcome, as are inspirations for new plugins and MCPs that could improve synergy and coverage. See CONTRIBUTING.md for full guidelines.

  • Issues: use the issue templates. Bug reports, feature requests, and plugin ideas each have their own form, but please check open issues for duplicates first.

  • PRs: Fork, branch, and follow the PR template. Good contributions include better hook triggers, bug fixes, documentation, and performance improvements, and any novelties that improve experience or perforamnce.

  • New plugins: Please Open a plugin idea issue; always looking for plugins that have clear trigger events, often wrap specific MCPs, or provide obvious synergy with their sibling plugins. Above all, always looking to solve a concrete user pain for other Claude users.

troubleshooting

  • plugin not triggering? Verify the mcp server is running: claude mcp list

  • mcp not found? Install via npm: claude mcp add -- npx

  • hook errors? Check that node.js is available and CLAUDE_PLUGIN_ROOT is set.

license

MIT


The Consummation of Empire — Thomas Cole, 1836

"alea iacta est" | the die is cast

— Julius Caesar, crossing the Rubicon

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
claude-emporium 怎么安装?

在客户端按下面的安装命令接入:npx -y claude-emporium。对应的包名是 claude-emporium。装完在客户端配置里引用同一个包或端点。

claude-emporium 怎么在 Claude Desktop、Cursor、Cline 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

claude-emporium 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

claude-emporium 需要 API Key 吗?

资料没有标注鉴权方式,是否需要密钥取决于服务本身,接入前按官方文档确认。

claude-emporium 支持哪些客户端?

资料标注支持 Claude Desktop、Cursor、Cline、Claude Code;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

claude-emporium 能提供哪些能力?

资料显示能力标签包括 claude-emporium、ai、ai-agents、anthropic、automation。具体工具清单以官方文档为准。

claude-emporium 需要什么运行环境?

运行时为 Cross-platform;属于本地进程,需要在本机装好运行环境。

claude-emporium 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-04-08。