ClawSouls MCP Server
Provides tools for AI agent persona management, safety verification, and efficient swarm memory synchronization across multiple agents. It enables users to search and install personas while utilizing a token-efficient Git-based workflow for maintaining shared project context.
README
ClawSouls MCP Server
AI agent persona management, safety verification, and memory tools via Model Context Protocol.
9 tools for Claude Code, OpenClaw, and any MCP-compatible client.
Quick Install (Claude Code)
# Install the ClawSouls plugin (includes MCP server)
/plugin marketplace add https://github.com/clawsouls/clawsouls-claude-code-plugin
/plugin install clawsouls
/reload-plugins
Or add directly to your .mcp.json:
{
"mcpServers": {
"clawsouls": {
"command": "npx",
"args": ["-y", "clawsouls-mcp@latest"]
}
}
}
Tools
🎭 Persona Management
| Tool | Description |
|---|---|
soul_search |
Search AI agent personas by keyword, category, or tag |
soul_get |
Get detailed info about a specific persona |
soul_install |
Download a persona and generate CLAUDE.md |
🔍 Safety & Integrity
| Tool | Description |
|---|---|
soul_scan |
SoulScan — verify persona safety against 53 patterns (A+ to F grade) |
soul_rollback_check |
Detect persona drift by comparing current vs. baseline files |
🧠 Swarm Memory
| Tool | Description |
|---|---|
memory_search |
TF-IDF + BM25 ranked search across MEMORY.md + memory/*.md |
memory_detail |
Fetch full content of a specific memory section (3-layer step 2) |
memory_status |
Show memory file inventory, sizes, and git status |
memory_sync |
Git-based multi-agent memory sync (init/push/pull/status) |
Memory Search
TF-IDF + BM25 Ranking (Default — Free)
memory_search query="SDK version fix"
Returns a compact index (~50 tokens per result) ranked by relevance:
| # | Location | Section | Score |
|---|-----------------------|------------------|-------|
| 1 | memory/2026-03-31.md:5 | SDK 버전 문제 해결 | 2.41 |
| 2 | MEMORY.md:42 | Troubleshooting | 1.87 |
Enhanced Mode (More tokens, more context)
memory_search query="SDK version fix" enhanced=true
Returns full snippets with score visualization for top results.
3-Layer Workflow (Token Efficient)
Step 1: memory_search query="bug fix" → compact index with scores
Step 2: memory_detail file="memory/2026-03-31.md" line=5 → full section
Step 3: (optional) memory_search enhanced=true → deep dive
~10x token savings compared to loading all memory files.
Swarm Memory Sync
Share memory across multiple agents via Git:
# Initialize (one time)
memory_sync action=init repo_url=git@github.com:user/agent-memory.git
# Push local changes
memory_sync action=push agent_name=brad
# Pull from other agents
memory_sync action=pull
# Check sync status
memory_sync action=status
Compatible Folder Structure
Works with OpenClaw's memory layout:
MEMORY.md # Long-term curated memory
memory/
topic-*.md # Project-specific status/decisions/history
YYYY-MM-DD.md # Daily logs
Platforms
| Platform | Integration |
|---|---|
| Claude Code | Plugin + MCP — /clawsouls:* commands |
| OpenClaw | Native SOUL.md support — always-on AI partner |
| Cursor / Windsurf | MCP server via .mcp.json |
| Any MCP Client | npx -y clawsouls-mcp@latest |
Links
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。