MELRA
Overview
MELRA is an open-source local runtime that governs the tool calls an AI agent routes through it over MCP. It checks each call against policy, asks for scoped approval when needed, verifies declared evidence, writes redacted hash-linked receipts, and parks uncertain mutations as recovery_required instead of replaying them. Adapters cover files, terminal, browser, computer control, HTTP and memory. Alpha software; requires Node.js 22.13+.
Key Features
- Policy checks at planning and execution time with task-scoped approvals
- Hash-linked redacted receipts and evidence-based verification
- Effect adapters for files, terminal, browser, computer control, HTTP and memory
- Works with MCP clients such as Claude Code, Cursor and VS Code over local transports
Best For
Agent builders who want explicit policy, approvals and audit receipts around local MCP tool execution.
Pros & Cons
Cons
- Alpha software; APIs may change before 1.0
- Only governs effects routed through MELRA; a second native tool path is not contained
- OS-level isolation is partly shipped and left to the operator
- Benchmark results are self-authored; database and cloud adapters are only planned
Pulse Verdict
Editorial review by SEODataPulse · Editorial rating: 3/5 · Updated
This is an editorial assessment, not an average of customer ratings.
“A documented, open-source governance layer for MCP agent tool calls with approvals, receipts and recovery handling. Still alpha, and it only governs actions routed through it, so treat it as one control among several.”
Pricing
Free and open source (Apache-2.0); self-hosted
Pricing changes often — confirm current plans on the official site.