LLMBrake

LLMBrake is a TypeScript library that sets per-customer monthly dollar cost caps and switches models for AI applications.

LLMBrake is a small TypeScript library that wraps OpenAI-compatible chat completion calls to track costs per customer. It computes call costs in USD from provider token usage and maintains a monthly spend total in UTC calendar months. When a customer approaches their budget limit, it automatically switches to a cheaper model, and it throws a budget exceeded error once the limit is reached. The library operates directly inside your application using your own API keys without acting as a proxy or reading prompt content. It supports streaming responses, custom price tables, and pluggable storage backends like Redis or memory. An optional hosted layer adds shared counters and a management dashboard.

LLMBrake at a glance

Best for: Software engineer
Runs on: Node, API, self-hosted
Pricing: Library is free and open source under the MIT license; LLMBrake Cloud is an optional paid layer.
Released: November 2026
Category: AI coding tool
Website: github.com/llmbrake/llmbrake

Key features

Tracks monthly customer spending in USD
Switches to cheaper models near budget limits
Blocks requests when budgets are exhausted
Supports streaming responses with token usage
Works with OpenAI-compatible endpoints
Offers pluggable storage for multi-server setups
Runs inside your application with direct provider calls

Use it for

Enforcing monthly spending limits per AI app customer
Downgrading expensive model requests near budget caps
Preventing unexpected API bills from runaway usage
Tracking token costs without proxying network traffic

Alternatives to LLMBrake

LiteLLM — One OpenAI-format API for 100+ LLM providers
FreeLLMAPI — Accesses hundreds of free LLM models and billions of tokens through one API key.
LangChain4j — Integrates LLMs and vector stores directly into Java applications.
Freebuff — Access multiple coding AI models for free without subscriptions.
SmallCode — Runs fully local with budget-managed context and a forgiving parser for small models.
MTPLX — It runs local models twice as fast via native MTP speculative decoding.
Octopus — Connects multiple LLM providers with automatic failover and protocol conversion.
Tokscale — Track and compare token usage across multiple AI coding agents.
oMLX — It uses paged SSD KV caching to reduce agent TTFT to under 5 seconds.
llm-d — Deploys production-grade LLM serving on existing hardware clusters.
Tallyhook Collector — Calculates exact AI coding agent costs locally and groups them for client invoicing.
ScaledThought — Access open-weight models on an OpenAI-compatible API set up entirely by an AI agent.

All LLMBrake alternatives