Awesome MCP › Agent Platforms

PV-Bhat/vibe-check-mcp-server

⭐ 502 TypeScript added to this list on 2025-05-14 repository created 2025-03-22

Vibe Check MCP is a specialized server designed to enhance the reliability and safety of AI agent workflows by implementing strategic pattern interrupts to prevent cascading errors. It functions as a meta-mentor layer that keeps AI agents aligned, reflective, and safe by stopping them from over-engineering and encouraging them to follow the minimal viable path first. The server uses a tool call named "Vibe Check" integrated with LearnLM 1.5 Pro (Gemini API), which is fine-tuned for pedagogy and metacognition to improve complex workflow strategies and prevent tunnel vision errors. The project is based on research demonstrating that agents using Vibe Check significantly improve success rates (from 27% to 54%) and reduce harmful actions by half (from 83% to 42%). It employs Chain-Pattern Interrupts (CPI) to interrupt pattern inertia and prevent Reasoning Lock-In (RLI), effectively acting as a rubber-duck debugger for large language models (LLMs). This approach allows agents to pause and reassess when risk levels spike, using metacognitive signals such as traits, uncertainty, and risk scores to trigger interventions before continuing. Vibe Check MCP is implemented as a lightweight server compatible with Anthropic's Model Context Protocol (MCP) and supports multiple clients including Claude, Claude Code, Cursor, Windsurf, and VSCode. It is easy to install and run using Node.js (version 20 or higher) and can be deployed locally or accessed via hosted runtimes like Smithery.ai. The project is well-recognized in the MCP community, featured on popular MCP registries and platforms, and has a strong user base with tens of thousands of installs. The server architecture allows it to run alongside existing agent workflows, emitting signals that inform the CPI mechanism to enforce intervention policies. This ensures that AI agents remain on track, avoid over-complexity, and maintain high levels of oversight and reliability. The project is open-source under the MIT license and encourages community contributions. Keywords: Model Context Protocol, MCP, AI agent oversight, Chain-Pattern Interrupts, CPI, Reasoning Lock-In, RLI, metacognition, pedagogy, LearnLM 1.5 Pro, Gemini API, AI workflow safety, minimal viable path, AI meta-mentor, Vibe Check, Anthropic MCP, AI alignment, AI reflection, AI safety, Node.js, Smithery.ai, open-source, MIT license.

https://github.com/PV-Bhat/vibe-check-mcp-server

agent-basedagentic-aiagentic-workflowai-agent-oversightai-agentsai-alignmentai-meta-mentorai-reflectionai-safetyai-workflow-safetyanthropic-mcpautomationchain-pattern-interruptscpierror-handlinggemini-apilearnlm-1.5-promcpmcp-servermetacognitionminimal-viable-pathmit-licensemodel-context-protocolmodel-context-protocol-serversnode.jsopen-sourcepedagogyreasoning-lock-inrlismithery.aivibe-checkvibe-codingvibecodingworkflow-automation