Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - builderall

#1
Fair skepticism — giving an AI write access to a firewall does sound alarming on the surface.

A few clarifications:

The MCP server connects via a restricted API key scoped only to firmware and diagnostics endpoints. It cannot touch firewall rules, VPN, DHCP, or any other config.
Write tools require explicit confirmation — Claude asks before running any update, upgrade, or reboot. It's more like a structured CLI than an autopilot.
OPNSENSE_READ_ONLY=true disables all write tools entirely if you only want monitoring.
Every API call is logged under System > Log Files > Audit.
As for the upgrade problem — glad you haven't hit it. It typically surfaces during major version upgrades (branch changes), not minor updates. The pkg ABI mismatch after a base/kernel upgrade is a known issue, just not one everyone encounters.

The SSH script exists precisely for people who want zero AI involvement and full control — it's a standalone Python script with no external dependencies.

Thank you for your comments.
#2
I've been working on tools to make OPNsense upgrades more reliable and recently wrote up the experience:

I gave Claude access to my OPNsense firewall — here's what happened

Two tools in one project:

A stateful Python upgrade script that runs over SSH — handles the pkg ABI mismatch after base/kernel upgrades and auto-resumes after reboots
A Claude MCP server that connects Claude Code to the OPNsense REST API for conversational firewall management
Tested on a live 26.1.2 → 26.1.3 upgrade this week, with two bugs found and fixed mid-session.

Code: https://github.com/builderall/opnsense-upgrade

Happy to answer questions or take feedback.