runcycles/cycles-openclaw-budget-guard
OpenClaw plugin for AI agent runtime control — enforce LLM cost limits, tool call caps, action permissions, and audit trails on OpenClaw agents before execution
Cycles budget and action guard for OpenClaw agents
- TypeScript99.1%
- Shell0.9%
1 Review
Cycles OpenClaw Budget Guard is a strong, practical plugin with a clear production use case: controlling autonomous agent spend before model or tool calls execute. The README is unusually good, explaining real failure modes like runaway retries, tool loops, noisy neighbors, and missing session-level visibility. The implementation backs that up with model/tool reservations, cleanup at agent end, configurable fail-closed behavior, budget-aware prompt hints, tool allow/block lists, call limits, metrics, dry-run mode, and broad tests around edge cases.
The main critique is polish rather than direction. Version signals appear inconsistent across README examples, code comments, and package.json, which can weaken user confidence. The repo would also benefit from obvious local test and coverage scripts, and CI linting if maintainers can support it. hooks.ts is doing a lot, so a future state/coordinator split may help maintainability.
Overall, this is a well-scoped, operator-aware project with strong documentation and serious production instincts.
