# AI Fieldcraft

> Operational field manuals for running AI in practice: local models on Mac and PC (Ollama, LM Studio, MLX), cloud workflows, AI agents, harnesses, agent loops, AGENTS.md briefing files, and agentic engineering. Exact commands, exact configs, model-by-RAM tables, working code, documented failure modes. Free manuals below; paid deep-dive manuals, a complete library, and a membership (The Ops Room) at https://aifieldcraft.com/.

## Free field manuals
- [FM-01: Run Your First Local AI Agent in 20 Minutes](https://aifieldcraft.com/guides/fm-01-first-local-agent.html): Ollama install, model pull, and a complete plan→act→observe agent loop in ~80 lines of Python with tool safety rules.
- [FM-02: Ollama on a Mac — The Complete Field Manual](https://aifieldcraft.com/guides/fm-02-ollama-mac.html): Apple Silicon unified memory, the model-by-RAM table (8/16/32/64/128 GB), OLLAMA_CONTEXT_LENGTH and keep-alive fixes, serving models to a network, MLX vs Ollama.
- [FM-03: The AGENTS.md Template That Fixes 80% of Agent Failures](https://aifieldcraft.com/guides/fm-03-agents-md.html): why coding agents fail, the complete copy-paste AGENTS.md template, monorepo nesting, and anti-patterns.
- [FM-04: Local vs Cloud AI — The Break-Even Field Manual](https://aifieldcraft.com/guides/fm-04-local-vs-cloud.html): the break-even formula, worked cost examples (used GPU vs API spend, the Mac zero-cost case), and the local/cloud/hybrid decision table.
- [FM-05: The Model Picker](https://aifieldcraft.com/guides/fm-05-model-picker.html): which local model for which job by hardware class, the VRAM fit rule (~0.6–0.7 GB per billion parameters at Q4), quantization doctrine, and a 10-minute personal evaluation drill.

## Services
- [Field Support](https://aifieldcraft.com/setup/): fixed-price done-with-you engagement — a working local/hybrid AI agent setup on your hardware in one week, with an operator's runbook.

## Contact
- corrections@aifieldcraft.com (defect reports), contact@aifieldcraft.com