Show HN: DashClaw – intercept and audit AI agent decisions before they execute (opens in new tab)

(github.com)

2 pointsucsandman13d ago2 comments

Hi HN,

I’ve been experimenting with autonomous agents for the past year, and I kept running into the same uncomfortable problem:

Agents were making decisions I couldn’t fully see or justify.

They could call tools, trigger actions, and make assumptions based on incomplete context. Once systems start touching real infrastructure, that becomes pretty scary.

So I built DashClaw.

DashClaw sits between an agent and the tools it wants to use. Instead of executing actions directly, agents call DashClaw first.

DashClaw can:

• evaluate the decision • apply policy rules • require approval • log reasoning and assumptions • record the final outcome

The idea is to create a *decision trail* for agent systems so you can answer:

Why did the agent think this was okay? What information did it rely on? Who approved the action?

The project includes:

• Node and Python SDKs • a decision ledger for agent actions • policy guardrails before execution • a mission control dashboard for fleet activity • a self-hosted architecture

It’s completely open source and designed to be lightweight enough to run locally with agent frameworks.

I'm still figuring out what the right abstraction layer is for this kind of infrastructure, so I’d love feedback from people building agents.

Repo: https://github.com/ucsandman/DashClaw

2 comments

ucsandmanOP13d ago

One design question I ran into was where governance should actually live in an agent stack.

If you put guardrails inside the prompt, the model can ignore them.

If you put them inside the agent framework, they can be bypassed.

DashClaw tries to solve this by intercepting actions instead of prompts. The agent can reason however it wants, but execution goes through a policy layer.

Curious how others are approaching this.

dreadpirates9d ago

Good framing on where governance should live. Intercepting actions outside the prompt and outside the framework is the right call, the model and the orchestrator are both untrusted surfaces. We took the same approach for spend specifically at nornr.com: agent requests a mandate before committing money, policy decides approve/queue/block at the infrastructure layer, every decision gets a signed receipt. Curious whether DashClaw distinguishes between 'read' actions and 'actions that cost money' — that's where we found the policy logic gets interesting.

j / k navigate · click thread line to collapse