alt Hacker NewsShow HN: Claw Patrol, a security firewall for agents
At Deno we've been using OpenClaw and other agents increasingly for addressing production problems in Deno Deploy - when a PagerDuty alert fires, the agent starts researching the cause and making fixes.
In order to do this, the agent needs access to real production systems - postgres, kubernetes, gcp, clickhouse, github, etc. But this is dangerous to say the least - we want destructive actions to be reviewed by other LLMs, approved by humans, and logged appropriately.
Claw Patrol terminates TCP connections over WireGuard or Tailscale, then parses application protocols (eg http, postgres, ssh) to apply rules that allow you to deny/allow requests.
There are a few projects that sit as a proxy in front of agents to do secret injection or apply various guardrails, but none met our needs (LLM gateways, MCP proxies, sandboxes), particularly the need to handle low-level protocols, or handle complex real world situations like tunneling postgres through k8s.
Written in Go, configured in HCL, MIT licensed. Happy to answer any questions.
Comments
I think this sounds very cool! It sounds similar to Agent Vault (github.com/Infisical/agent-vault) but with an added feature of having security policies for denial/human-in-the-loop of traffic based on the contents of requests?
The nice thing about Agent Vault is the encryption of credentials and other ways they handle making sure those don't leak from storage. I suppose you could potentially wrap the two in layers as well (agent -> Claw Patrol -> Agent Vault -> external network)
EDIT: looking at some of the comments, it sounds like Claw Patrol can work with protocols beyond HTTP/S, so potentially covers more surface area than AV
The product looks great and I'm really interested in trying it out. Very cool, congrats on shipping! Also...as a parent of young kids: this name made me laugh out loud. The OG image on the marketing site is a fun easter egg.
For those here without young kids in their life: https://en.wikipedia.org/wiki/Paw_Patrol
Nice work shipping this.
Disclosure: author of a related tool here. I have create agent-vault-proxy for a very similar reason. It also can help keep credentials out of the agent process. The agent gets a placeholder, the proxy swaps in the real secret in transit.
I read them as complementary: action firewall in front, credential broker behind. https://github.com/inflightsec/agent-vault-proxy
Interesting project. I like the implementation, congratulations on shipping!
This is a really cool library to look at even if you aren't running openclaw directly.
Lots of good concepts to seek inspiration from.
1. process-scoped egress policy
2. policy-as-code
3. explicit approval classes
4. normalized network/ guardrail receipts.
5. structured guardrail outcomes
6. centralized decision rules
This is very interesting. I build something like this but native to claude code and something that focus on just logging the violation. My question is if you are terminating a process with in the workflow will that about all other things that executed before. anyway would love your feed back on this https://github.com/varmabudharaju/agent-pd
Seems like a more general solution to a Tesla API Firewall that i was thinking about. My idea was to use some kind of gateway/firewall LLM to check commands that another agent would send to the Tesla API.
How will credentials be injected via Gateway for each user ? If we have 5 users with one gateway, how it knows whose github credential to inject ?
Neat! Reading the docs - it's default-allow and ships with no rules? Any plans for a default rule set?
I did this
really interesting work! i am curious how you handle rule configuration for different protocols such as Postgres or ssh. Thanks for open-sourcing it under MIT.
[flagged]
[flagged]
[flagged]
Great name by the way, as someone who's been forced to watch the show by cousins.