sub-agents-mcp
Verified Safeby shinpr
Overview
Acts as an MCP server to enable Claude Code-style sub-agents in any MCP-compatible tool like Cursor, facilitating delegation of specialized AI tasks defined in markdown files.
Installation
npx -y sub-agents-mcpEnvironment Variables
- AGENTS_DIR
- AGENT_TYPE
Security Notes
The server implements strong input validation and path traversal prevention for agent names, prompts, and session IDs. Session files are stored with restrictive permissions (0o600). It avoids hardcoded secrets by relying on environment variables (e.g., CLI_API_KEY). No 'eval' or obfuscation found. The primary security consideration is the reliance on external CLI tools (`cursor-agent` or `claude`) to execute sub-agents, which themselves require explicit shell command allowlisting by the user, making their security a responsibility of the user's environment configuration.
Similar Servers
rmcp
Serves as an AI assistant backend to perform comprehensive statistical analysis, econometric modeling, machine learning, time series analysis, and data science tasks using R through natural language conversations.
company-docs-mcp
Transforms organizational documentation into an AI-powered knowledge base for semantic search, Q&A via chat interface, Claude Desktop, and Slack integration.
karma
Provides a plug-and-play Knowledge Graph service via the Model Context Protocol (MCP), allowing AI agents to store, retrieve, and manage structured and unstructured knowledge (entities, relations, observations).
fastmcp
Build fast, declarative, and type-safe servers implementing the Model Context Protocol (MCP) in Dart.