Mindburn-Labs/helm-ai-kernel

View on GitHub ↗

Fail-closed execution firewall for AI agents: quarantine MCP tools, proxy OpenAI-compatible requests, emit signed receipts, and verify EvidencePacks offline.

56 ★3 forksGoUpdated 18d ago

What you need to know

An MCP server that provides a secure AI kernel/sandbox for running untrusted AI code, managing workflows, sessions, and resource limits for agent execution.

Install

Deploy the kernel server (container-based sandbox) and configure the MCP endpoint.

Usage

  • Run AI-generated code inside an isolated sandbox with resource controls via tools.

Key features

  • secure sandbox
  • AI code execution
  • session management
  • resource limits
  • workflow controls

Best for

Running untrusted AI-generated code safely.

Platforms: macOS · Linux

Reviewed 2026-08-11

Topics

agent-securityai-agentsai-securityclideveloper-toolsdevsecopsevidencepackexecution-boundarygollm-securitymcpmodel-context-protocoloffline-verificationopenai-compatiblepolicy-engineproofgraphsandboxself-hostedsigned-receiptszero-trust
Stars
56★
Forks
3
Language
Go
License
Apache-2.0
Created
2026-02-15
Last push
2026-08-17