The io.github.ejentum/self-inspect-mcp server is an MCP server that lets you send a thought and receive one “metathought” prompting an agent to inspect its own assumptions. It is described as keyless and provides 1 tool.
🛠️ Key Features
- Sends a thought and returns a metathought for self-inspection
- Keyless design
- Implements Model Context Protocol (MCP)
🚀 Use Cases
- Agentic workflows that require metacognition and self-reflection
- Reasoning loops focused on inspecting assumptions
- Self-checking behavior for LLM/AI agents
⚡ Developer Benefits
- Small, focused MCP integration with toolCount: 1
- Topic-aligned support for agentic-ai, agents, and ai-agents
- Built around metacognition and self-reflection
⚠️ Limitations
- Only 1 tool is provided, limiting scope to the described thought-to-metathought behavior