BricksLLM is an open-source, enterprise-grade AI gateway for putting large language models into production, with cost and rate limits, access control and monitoring per API key. Built by a Y Combinator-backed team and written in Go, it has about 1,200 stars on GitHub and is licensed under MIT.
When LLM features move from prototype to production, teams need the controls a normal API gateway provides, adapted to AI: who may call which model, how much they may spend, and what happens when a provider fails. BricksLLM sits between your applications and providers such as OpenAI, Azure OpenAI, Anthropic and vLLM, enforcing per-key, per-user and per-environment cost and rate limits, tracking usage in fine detail, and improving reliability with failovers, retries and caching. It can also detect and redact PII in requests. A managed cloud version with a dashboard is offered at trybricks.ai for teams that prefer not to self-host.
The gateway is provider-agnostic at the API layer, so applications keep a single integration while the team changes models, adds providers or splits traffic behind the scenes. Policies, keys and usage records are managed through its admin interfaces and stored in PostgreSQL, which makes BricksLLM a fit for the same operational habits teams already apply to the rest of their infrastructure.
BricksLLM is freemium: the gateway itself is free and open source under the MIT license for self-hosting, while the managed cloud version at trybricks.ai offers a hosted dashboard with paid plans after a free start.
BricksLLM is for platform and product teams running LLMs in production who need spend control, key management and observability across multiple model providers.