CURSED-ME / loopers-oss

What is loopers-oss?

Loopers is a reverse proxy written in Go that sits between your application and AI providers like OpenAI, Anthropic, and Gemini. It enforces hard budget limits before requests reach the provider--not after. You set spending caps across five time windows (minute, hourly, daily, weekly, monthly), and the proxy returns HTTP 429 if a request would exceed your budget. It handles streaming responses too, cutting off mid-stream if token counts push you over limit. Deployment runs via Docker or Kubernetes Helm charts, with Redis as the backing store for atomic budget tracking.

Why is it gaining traction?

The hook is simple: runaway AI agents can burn thousands of dollars in minutes. Loopers promises to kill that problem at the infrastructure level, not through alerts or dashboards but through a genuine kill-switch. It claims sub-2ms overhead, atomic Lua-script guarantees against race conditions, and a fail-closed design that blocks all requests if Redis goes down. The streaming mid-stream cutoff is particularly notable for long-running agent tasks.

Who should use this?

Backend engineers building autonomous AI agents or multi-step workflows that call LLMs repeatedly. Startups and small teams who want hard budget guardrails without building custom middleware. Anyone running AI agents on shared API keys where one runaway loop could blow through a monthly budget. Not ideal for one-off scripts or casual experimentation where simple API key quotas suffice.

Verdict

The architecture is sound--test coverage exists, Lua scripts handle atomic operations, and the CLI makes key management practical. But 19 stars and a credibility score of 0.7% mean this is a young project still proving itself. The demo works as documented, which is a good sign. If you have budget-sensitive AI workloads, it is worth spinning up the docker-compose demo to evaluate. Just do not deploy to production without running your own load tests first--community support is thin at this stage.