PV-Bhat/vibe-check-mcp-server

Overview

An MCP server written in TypeScript that provides mentor-like feedback to AI agents during complex workflows. It intercepts agent reasoning to prevent tunnel vision, over-engineering, and reasoning lock-in, helping agents stay focused on the true objective.

Best for

Best for
Developers building complex, long-horizon AI agents that need guardrails against over-engineering and reasoning lock-in.

Use cases

• Preventing AI agents from over-engineering code solutions
• Keeping agents on track during ambiguous or multi-step tasks
• Reducing reasoning lock-in for high-risk agent workflows

How to use

Install

npx -y @pv-bhat/vibe-check-mcp start --stdio

Tested with

Claude Desktop, Cursor, Windsurf, Continue, VS Code, ChatGPT

Example client config

{\n  "mcpServers": {\n    "vibe-check-mcp": {\n      "command": "npx",\n      "args": ["-y", "@pv-bhat/vibe-check-mcp", "start", "--stdio"]\n    }\n  }\n}

Notes

An MCP server written in TypeScript that provides mentor-like feedback to AI agents during complex workflows. It intercepts agent reasoning to prevent tunnel vision, over-engineering, and reasoning lock-in, helping agents stay focused on the true objective.

487 stars on GitHub. Last updated 2026-05-28. Licensed MIT.

Use cases

• Preventing AI agents from over-engineering code solutions
• Keeping agents on track during ambiguous or multi-step tasks
• Reducing reasoning lock-in for high-risk agent workflows

Pros

• Open-source with a growing community (487 stars on GitHub)
• Addresses a common pain point: agent tunnel vision in long-horizon tasks
• Lightweight integration via the MCP protocol, works with existing agent frameworks

Cons

• Requires the agent to be MCP-compatible, limiting direct use with many current LLM APIs
• Feedback quality depends on the underlying prompt or model used for mentor logic
• May add latency to agent reasoning cycles if called too frequently

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.