Every token counts.
ScroogeLLM is a VS Code extension that sits between you and your LLM provider. It shows you exactly what every call costs — then optimizes it automatically.
No spam. One email when we launch.
The Problem
Most developers have no idea what each AI call actually costs. ScroogeLLM makes every dollar visible — then recovers what you can.
Features
From real-time cost tracking to automatic optimization — all running locally on your machine.
See exactly what every LLM call costs as it happens. Per-request, per-session, and cumulative totals right in your status bar.
FreeAutomatic context trimming removes redundant tokens before they hit the API. Same quality responses, fewer tokens billed.
Identical or near-identical prompts return cached results instantly. Zero cost, zero latency on repeat queries.
Not every task needs the most expensive model. ScroogeLLM suggests downgrades when a cheaper model can handle the request equally well.
Deterministic fake names replace real PII before it reaches the provider. Same fake name for the same input, stable across your session.
FreeEvery request logs raw cost vs. actual cost. Full audit trail you can inspect, export, and use to justify tooling budgets.
FreeAPI keys stored in your operating system's native keychain. Never plaintext, never transmitted. Your keys, your machine, period.
FreeCore optimizations work without paying a cent. Visibility, PII protection, and audit logging are free forever. Paid features add deeper savings.
FreeHow It Works
ScroogeLLM runs a lightweight proxy on localhost. Your AI tools talk to the proxy. The proxy talks to the provider. In between, it does its work.
Install from the VS Code Marketplace. Point your LLM tools at
localhost. Done.
Every request flows through the proxy. Prompts are compressed, PII is scrubbed, and responses are cached automatically.
See real-time costs in VS Code. Review your audit trail. Watch the savings accumulate, request by request.
graph TD
subgraph localhost["🖥 Your Machine (localhost only)"]
A["VS Code / AI Tool"] --> proxy
subgraph proxy["ScroogeLLM Proxy"]
direction LR
subgraph row1[" "]
direction LR
B["compress"] --> C["anonymize"] --> D["cache"]
end
subgraph row2[" "]
direction LR
E["route"] --> F["log"] --> G["audit"]
end
row1 --> row2
end
end
proxy --> H["☁️ LLM Provider API\n(OpenAI, Anthropic, etc.)"]
style localhost fill:#0d1117,stroke:#1a3a2a,color:#8b9a8b
style proxy fill:#0a1f0a,stroke:#22c55e,color:#4ade80
style row1 fill:transparent,stroke:none
style row2 fill:transparent,stroke:none
style A fill:#1a1a2e,stroke:#6c9cfc,color:#a0b4e0
style B fill:#132613,stroke:#22c55e,color:#4ade80
style C fill:#132613,stroke:#22c55e,color:#4ade80
style D fill:#132613,stroke:#22c55e,color:#4ade80
style E fill:#132613,stroke:#22c55e,color:#4ade80
style F fill:#132613,stroke:#22c55e,color:#4ade80
style G fill:#132613,stroke:#22c55e,color:#4ade80
style H fill:#1a1a2e,stroke:#6c9cfc,color:#a0b4e0
Privacy & Security
ScroogeLLM is designed for developers who take data seriously. No cloud, no accounts, no telemetry.
The proxy binds to 127.0.0.1 by default. No remote exposure without your explicit opt-in.
API keys live in macOS Keychain, Windows Credential Locker, or Linux Secret Service. Never in plaintext files.
No usage data, no analytics, no phone-home. We never see your prompts, your code, or your API keys.
Real names, emails, and identifiers are replaced with stable fakes before data leaves your machine. Same input, same fake, every time.
Get Early Access
ScroogeLLM is coming to the VS Code Marketplace. Leave your email and be the first to know.
Free forever tier included. No credit card needed.