Hey HN, we built Bulwark (https://getbulwark.ai) because the current permission model for coding agents doesn't scale. Having to constantly check back in to approve or reject commands defeats the purpose. One agent, one terminal, one human watching is fine, but to get the most out of these agents, we need to run multiple sessions across laptops, servers, and cloud dev environments with the same permission set.

Bulwark is a centralized permission layer for coding agents. Define a single set of policies that are enforced across all sessions and devices, approve uncommon commands from a single dashboard with a full audit trail. Let your agents run at full speed and be confident that they won't wreak havoc on your machines and integrations!

Works with Claude Code and Amp now, with more integrations coming.