Open-source repositories, long-form handbooks, and technical writing at the intersection of LLMs, agentic systems, and developer tools — built in the open, shipped in public.
Vibe Engines is a one-person workshop publishing open-source tools, technical writing, and long-form handbooks on the parts of AI that actually ship — agent loops, LLM orchestration, prompt engineering under real constraints, and the glue between them.
The operating metaphor is a workshop, not a blog and not a lecture hall. Everything here is made by someone doing the work, for people doing the work. Every repository ships something that runs. Every handbook is written from the floor — specific, opinionated, and honest about what didn't work.
Some of the tools will get used. Some won't. That's the point of running a workshop in public — you get to find out, and so does everyone else.
If you're wiring up an LLM product on weekends and need real patterns — evals that catch regressions, caching that actually saves money, agent loops that don't spiral — the handbooks and repos here are written for that pace.
Stitching a stack together as a team of one means every dependency has to pull its weight. The tooling here is small, legible, and built to be forked — not framework lock-in dressed up as a library.
For backend and full-stack engineers moving into agentic systems, LLM infrastructure, and Claude / MCP workflows. Deep on the mechanics, light on the hype, and always with working code attached.
Full-length field manuals on agentic AI, LLM systems, and engineering craft. Free to read. Written from the workshop floor.
Browse handbooks →Tools, repositories, and small utilities — MCP servers, agent scaffolds, and Claude API helpers. All shipped in public on GitHub.
See projects →Shorter pieces on what's actually working — patterns from production LLM systems, notes on agent design, and occasional postmortems.
Read writing →The most useful thing here is usually the next handbook. Start there.
Go to the handbook library →