About Gurgl
Gurgl is an MCP server published by nurbanasaurus in the Security category: local-first egress hygiene for MCP servers: capture, diff, and allowlist what your AI-agent tools contact on the network.
The repository has 1 stars and 1 forks, with the last commit 3 months ago. Recent commit activity is the strongest signal that reported issues will get attention.
Install
git clone https://github.com/nurbanasaurus/gurglThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Tools (5)
Socket.dev, npm audit, dependency scannersSupply-chain metadata: who published, what changed in the package. gurgl adds the behavioral half - what the code actually does on the wire.
Parameters not published.
Squid or another egress proxyThe same idea at a network chokepoint: `--format squid` emits the ACL.
Parameters not published.
[Anthropic sandbox-runtime]Enforces a network allowlist on the running agent. `gurgl allow --format sandbox-runtime` generates the domain list from observed behavior instead of guesswork. Observe with gurgl, enforce with the sandbox - the strongest combo.
Parameters not published.
[OpenSnitch] (Linux), Little Snitch / LuLu (macOS)Per-app firewalls that ask "allow this connection?" at runtime. gurgl answers the question they cannot: what does this tool *normally* contact? `gurgl allow <server> --format opensnitch` turns a clean baseline into rules, so the firewall only prompts on genuine anomalies.
Parameters not published.
mcp-scan (Invariant Labs) and other static scannersAn orthogonal axis. They inspect tool *descriptions and prompts* (tool poisoning, injection); gurgl inspects *runtime network behavior*. A server can pass one and fail the other - run both when vetting.
Parameters not published.
Ask AI
Ask AI about Gurgl
Powered by Claude · Grounded in docs
Security checks
6 checks have not been run yet.