Show HN: Melra – pause ambiguous MCP writes instead of blindly retrying

Category: ai-ml

Tags: mcp, ai-agents, execution-runtime, policy-enforcement, tool-calls, reliability

Score: 6.8/10 (Innovation: 7, Technical: 7, Documentation: 7, Utility: 6)

MELRA is a local execution runtime that sits between AI agents and MCP tool calls, enforcing policy, requesting approvals, and verifying evidence to prevent blind retries of ambiguous writes. It coordinates at-most-once dispatch with parked reconciliation for uncertain outcomes, providing a receipt for governed actions. The project addresses a real reliability gap in agentic tool use, especially for MCP-based agents.

Target audience: AI engineers and developers building MCP-based agents who need reliable, auditable tool execution

Repository: https://github.com/XAGI-Lab/melra · TypeScript · Apache-2.0 · 12 stars

View on Hacker News