Open-Source Local Firewall for MCP
Simple-Factor-2165 · reddit · 2026-07-14
A developer has created MCP-Shield, an open-source local firewall designed to protect AI agents' MCP tool calls.
The core pain point: once an agent gets shell or tool permissions, it can be tricked by a single prompt injection into executing unintended actions. Existing permission systems often rely on the client, so the author chose to build a local proxy one layer below the MCP transport layer.
Design goals:
- Automatically allow obviously safe tool calls;
- Block obviously dangerous commands;
- Pause suspicious operations, executing them only after the user confirms, modifies, or rejects them.
The author explicitly states this doesn't "solve" prompt injection; if an attacker operates entirely within tools you already trust, the proxy can only do so much. However, it provides a unified layer of protection for clients like Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, and Codex CLI.
Project commitments:
- 100% local execution;
- No cloud services, no accounts, no telemetry;
- MIT open-source license;
- Approval interface is a local web UI.
More from coding & agent
- TMLR paper: enforced thinking doesn't reliably make LLM agents disclose key info to users — SharonYixuanLi · 2026-10-27
- Resin lets an agent build standalone LLM inference engines; vLLM and SGLang optimize DeepSeek kernels differently — AccBalanced · 2026-10-27
- How I let my cloud Grok Bot code on my Mac mini via Tailscale, no public ports — talkaboutdesign · 2026-10-27
- Context layout is a cost decision: one memory edit can bust your entire agent prompt cache — dscape · 2026-10-27
- Alchemy adds effect/Workflow support across Durable Objects, Workflows and Lambda Durable Functions — samgoodwin89 · 2026-10-27
- Alchemy's Lambda Web Functions run plain Node web servers on Lambda, no OpenNext needed — samgoodwin89 · 2026-10-27