Proxy Engineering — Stockyard"> Blog: LLM Infrastructure & Proxy Engineering — Stockyard
NEW: Stockyard now has 150 self-hosted tools. Browse the full catalog →

Blog

Building LLM infrastructure in the open. RSS

How I Cut My Cursor API Bill by 60% With a Local Proxy
Cursor sends more LLM requests than you think. Caching and model routing through a local proxy dropped my monthly API bill from $380 to $152.
Proxy-Only vs Full Platform: When to Use Each
Stockyard ships two binaries: an open-source proxy (Apache 2.0) and the full platform (BSL 1.1). Here's when to use each.
The True Cost of LLM APIs in 2026: 40+ Models Compared
We compiled pricing for 40+ models across 16 providers. GPT-4o, Claude, Gemini, Llama, Mistral, DeepSeek, Grok — what does 1 million tokens actually cost? Plus how to cut your bill 50–80%.
76 Modules, 400 Nanoseconds: Benchmarking an LLM Middleware Chain
We put 76 middleware modules between our app and OpenAI. PII redaction, cost caps, caching, rate limiting, audit logging — the entire chain adds 400ns. Here's how.
Why One Binary: Building an LLM Gateway with Go and SQLite
Why I chose Go over Python and SQLite over Postgres for an LLM proxy. The engineering tradeoffs behind shipping 150 tools as one static binary.
Why I Built Stockyard
I was building an app with the OpenAI API. Then I shipped it. Within the first week, three things went wrong.
Architecture Decisions Behind Stockyard
Why Go, why SQLite, why a single binary, and why 76 middleware modules instead of a plugin system.
Replacing 150 LLM Tools with One Binary
The LLM middleware ecosystem is fragmented across 134+ standalone tools. Here's how Stockyard consolidates them.
Explore: OpenAI proxy · vs LiteLLM · Best self-hosted proxy · What is an LLM proxy? · Install guide