# HELM AI Kernel

HELM AI Kernel is a fail-closed execution firewall for AI agents that intercepts tool calls, shell commands, and requests from agents such as Claude Code and Codex, deciding to allow, deny, or escalate each action and. It is open source and self-hosted, installable on Kubernetes (Helm).

Canonical page: https://ourlocalhost.com/app/4881/helm-ai-kernel

## Description

HELM AI Kernel is a fail-closed execution firewall for AI agents that intercepts tool calls, shell commands, and requests from agents such as Claude Code and Codex, deciding to allow, deny, or escalate each action and recording a verifiable signed receipt.

## Platform availability

Available on 1 of 10 platforms: Helm.

## Health score

55/100 (Fair), based on 4 of 5 factors.
- Maintained: 100 (last commit 14 days ago)
- Popular: 36 (61 GitHub stars)
- Easy to install: 25 (1 install platform, install notes)
- Security check: 20 (1 serious finding, 7 minor, CI, tests, OpenSSF Scorecard 6.7/10)

## Security check

1 serious finding: A GitHub workflow can be triggered by outsiders to run their code with the project's secrets.
Automated checks on commit 5a6e27a, 12 Nov 2026. Not an audit.

## Links

- GitHub: https://github.com/Mindburn-Labs/helm-ai-kernel
- Official website: https://helm.docs.mindburn.org/
- Kubernetes (Helm): https://artifacthub.io/packages/helm/mindburn-labs/helm-ai-kernel

## Common questions

### Which Helm chart failure categories should I investigate first?

The troubleshooting page lists lifecycle failures, networking/ingress issues, storage configuration errors, and component-specific operational hurdles.

### Can I verify EvidencePacks without external network access?

Yes, the documentation says it verifies EvidencePacks offline.

### Does it support OpenAI-compatible request traffic?

Yes, it proxies OpenAI-compatible requests.

---
Source: Our Local Host (https://ourlocalhost.com/app/4881/helm-ai-kernel), the self-hosted and open-source app directory.
