mcp-writ - Policy enforcement and sandboxing for MCP servers.

by
mcp-writ aims to support allowlist-based operation of local MCP servers by combining static analysis for policy drafting with runtime JSON-RPC checks and OS sandboxing.

Add a comment

Replies

Best
Maker
📌
The concept behind mcp-guard is to make the permissions of locally running MCP servers explicit and reviewable. The approach combines OS sandboxing with policy checks on tool calls and their arguments. It includes controls for tool access and filesystem permissions, audit logging, and draft policy generation to support configuration. The current scope is stdio MCP servers. Available restrictions depend on the operating system and the configured policy.
How did Astra change the scope or ambition of what you built?

I had left mcp-writ untouched for several months. The project aims to put explicit security boundaries around locally running MCP servers, combining OS sandboxing with policy checks on tool calls and their arguments.

Astra helped me revisit the project, incorporate up-to-date information, and get the implementation running. For a project that brings together operating-system restrictions, MCP protocol handling, and policy configuration, reaching a runnable state was a meaningful step.

It changed what I could realistically move forward with. A project that had been sitting dormant became something I could run, giving me a concrete implementation to support the concept I wanted to share.