tech.human/tap Model Context Protocol (MCP) Server
This MCP server provides credential isolation for AI agents using placeholder secrets, policy checks, optional human approval, and connector routing. It focuses on auditing request routing and connector behavior, debugging failed requests, and understanding approval flows for the TAP experience.
๐ ๏ธ Key Features
- Credential isolation for AI agents
- Placeholder secrets to prevent sharing sensitive data with agents
- Policy checks
- Optional human approval gating
- Connector routing for requests
๐ Use Cases
- Auditing request routing and connector behavior
- Debugging failed requests
- Understanding approval flows
- Improving connector-side request shaping
- Contributing fixes to the core TAP experience
โก Developer Benefits
- Easier debugging of routing and connector behavior
- Clear insight into approval flow behavior
- Helps trace and shape connector-side requests during fixes
โ ๏ธ Limitations
- Self-hosting requires responsibility for security of credentials and signing keys
- Documentation notes that TAP keeps secrets out of agents, but host hardening and correct policy handling are still required
Related Topics
- agent-security, ai-agents, credentials, llm-security, mcp, proxy, secrets-management