Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/scouzi1966/maclocal-api/claude-mdgit clone --depth 1 https://github.com/scouzi1966/maclocal-apiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/scouzi1966/maclocal-api/claude-md)<a href="https://agentmods.dev/instructions/scouzi1966/maclocal-api/claude-md"><img src="https://agentmods.dev/badge/instructions/scouzi1966/maclocal-api/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.04520 | $0.04520 |
| Opus 5 | $0.02260 | $0.02260 |
| Sonnet 5 | $0.00904 | $0.00904 |
| Haiku 4.5 | $0.00452 | $0.00452 |
Grade A, and why
maclocal-api CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 302 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude Code Reference for MacLocal API
Project Overview
AFM (Apple Foundation Models) is an OpenAI-compatible API server for local LLM inference on Apple Silicon. It supports two backends:
- MLX (
afm mlx): Uses mlx-swift-lm for MLX-format models from Hugging Face - Apple Foundation Models (
afmwithout subcommand): Uses Apple's on-device models (macOS 26+)
The server exposes /v1/chat/completions and /v1/models endpoints compatible with OpenAI API clients.
Project Structure
The package vends a maclocal-api aggregate library (AFMKit), the Vapor HTTP
layer (AFMServer), compatibility libraries, and the afm executable. Stable
provider contracts and runtimes come from the standalone AFMKit package at an
immutable revision. AFMServer remains local and adapts those provider facades
into OpenAI-compatible HTTP responses.
Sources/
├── AFMKit/ # maclocal-api aggregate facade and services
├── AFMServer/ # Vapor routes, HTTP lifecycle, metrics, files
├── AFMCLI/ # executable entry point and commands
├── AFMKitFoundationModels/ # macOS 26 Apple service compatibility
└── AFMKitFoundationModels27/ # facade over AFMKit macOS 27 products
vendor/
└── llama.cpp/ # WebUI source
Scripts/
├── check-afmkit-consumer-boundary.sh
└── build-from-scratch.sh # clean immutable consumer build
Provider source ownership
AFMKit is the only source of truth for MLX, mlx-c, mlx-swift-lm, xgrammar,
DwarfStar, and their AFM adaptations. maclocal-api consumes one exact AFMKit
version and must not vendor, fork, patch, or compile a second provider tree.
Use MACLOCAL_AFMKIT_WORKSPACE_PATH for reversible paired development, validate
both repositories, then publish and pin one new AFMKit version.
DwarfStar dependency boundary
AFMKit owns the canonical DwarfStar and xgrammar dependencies, their C/C++
bridges, and provider regressions. maclocal-api consumes those products only
through its resolved AFMKit dependency. Scripts/swiftpm-reliable.sh validates
and fingerprints that checkout so Xcode cannot reuse products compiled from a
different provider revision.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 302 lines · 4,520 tokens per session scan A e95944cb364a
maclocal-api CLAUDE.md is an instructions file published in the GitHub repository scouzi1966/maclocal-api (337 stars, last pushed today), licensed MIT. It adds 4,520 tokens to every session, about $0.0226 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
osaurus AGENTS.md
AGENTS.md instructions for osaurus-ai/osaurus, covering codex configuration - osaurus-staging, build & test, keychain tip (optional), osaurus release proof and pr reporting and model runtime non-negotiables.
Swift-Foundation-Models-Skills AGENTS.md
Instructions for sambitcreate/Swift-Foundation-Models-Skills, covering apple foundation models skills repository, skills overview, core session management, structured generation and tools & system integration.
apfel CLAUDE.md
Instructions for Arthur-Ficial/apfel, covering apfel - project instructions, the golden goal, core product (this is what apfel is), byproducts (useful, but not the pitch) and readme.md structure rule.
strix-halo-guide AGENTS.md
AGENTS.md instructions for hogeheer499-commits/strix-halo-guide, covering agents.md, core project lens, documentation style, do-not-invent rules and benchmark claim rules.
mectrics CLAUDE.md
Claude Code instructions for farukkamcici/mectrics, covering claude.md, absolute must-knows (see agents.md for detail), fast local loop (no xcode ui), build & run the app and notes.
Usage4Claude CLAUDE.md
Claude Code instructions for f-is-h/Usage4Claude, covering usage4claude — project overview, build & test, architecture map, key conventions and known pitfalls.