HN Showcase
Back to browse
RiskKernel, kill -9 an AI agent and resume it without paying twice

RiskKernel, kill -9 an AI agent and resume it without paying twice

by prashar32·Jun 10, 2026·2 points·4 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemSlick

Kill switch fires mid-loop at dollar ceiling before the bill lands.

Strengths
  • •Crash-resumable checkpoints mean kill -9 doesn't waste spent tokens
  • •Human-approval gates block side-effecting tool calls before they execute
  • •Single Go binary with OpenTelemetry export to existing observability stack
Weaknesses
  • •Agent guardrails category already has several funded competitors
  • •Proxy architecture adds latency layer between agent and provider
Category
AI/ML
Target Audience

Engineering teams running AI agents in production

Similar To

Helicone · Langfuse · Braintrust

Similar Projects

AI/ML●●Solid

RiskKernel – a kill switch and budgets for runaway AI agents

Deterministic Go governor halts runaway agents at budget with HTTP 402.

Big BrainShip It
prashar32
302d ago
Developer Tools●●Solid

Agent Firewall – Go proxy to kill LLM death spirals

Wire-protocol circuit breaker for agents when LangSmith costs too much.

Solve My ProblemShip It
wuweiaxin
213mo ago
AI/ML●●Solid

Kill-The-Backlog, self-hosted background agents

E2B sandbox isolation prevents agent chaos while streaming realtime previews before PR creation.

Ship ItBig Brain
jvaill
401mo ago