Doberman
Guard dog for your AI when it goes rogue.
Visit Website
September 1, 2026
PH launched today! Doberman, an open-source guard dog for AI coding agents
I think agent security is going to be a boring, unavoidable line item the way backups became one, and I wanted the open-source default to exist before the enterprise vendors define the category. Doberman sits between a coding agent and its tools, every tool call gets allow/authenticate/block from a local policy engine before it runs. Apache-2.0, local-first, pip install doberman-core.
The open-core bet: the runtime, policy engine, and adapters are free forever, and the paid layer later is compliance evidence and team controls. 19 contributors have merged PRs since it went public.
Live on Product Hunt today: https://www.producthunt.com/products/doberman
3 Comments
Solid positioning. Agent security and tool-call authorization are definitely heading toward that "boring, unavoidable line item" status as autonomous agents get more execution privileges.
Love the local-first approach with the policy engine sitting right in front of tool execution.
Just checked out the PH page and dropped an upvote/comment. Releasing open-source core with 19 PRs already merged is a serious validation signal. Congrats on shipping today!
PS: If you get a quick breather between PH notifications today, I’d love your eyes and technical feedback on LocalPatron over on the Build Board: https://www.indiehackers.com/product/localpatron?post=BZHMchLNDI4lxlFrw6xz
170+ PRs across 19 contributors in a single month is massive! That's serious open-source velocity right there.
Appreciate you taking a look at LocalPatron amidst the launch chaos today!