asymm-plugins

Mathematical superpowers for AI agents

9 tools. All local. Proven math. No cloud, no data leaving your machine.

---

# 1. Build the Rust binary
cd crates/asymmetrica-core && cargo build --release

# 2. Add to Claude Code
# Open Claude Code settings -> MCP Servers -> Add
# Point it at: ./plugin/plugin.json

That's it. Restart Claude Code and the tools are live.

---

| Tool | What it does | |------|--------------| | scan | Map any directory — files, sizes, types, age distribution, space hogs | | find_duplicates | Find duplicate files 9x faster than brute force (pre-filters 89% of candidates) | | find_stale | Classify files as active, dormant, or fossil — find reclaimable space | | build_graph | Build a persistent SQLite knowledge graph of your codebase | | query | SQL any previously scanned directory via the knowledge graph | | optimal_batch | Calculate the mathematically optimal batch size for any N items | | health_check | Detect hidden divergence between subsystems even when aggregates look fine | | cluster | Group files by similarity — find near-duplicates, related modules, semantic clusters | | smart_schedule | Generate optimal timing for spaced operations (32% less bursty than exponential backoff) |

---

A Rust binary does the heavy computation — quaternion clustering, digital root filtering, Williams-optimal batching. A Node.js MCP bridge exposes it to Claude Code as 9 tools. Everything runs on your machine; no data ever leaves.

---

cd crates/asymmetrica-core
cargo build --release
# Binary lands at: target/release/asymmetrica

Requires Rust 1.75+. No other dependencies.

---

Real workflows, copy-paste ready.

| Cookbook | What it teaches | |----------|-----------------| | 01 — First Look at a Codebase | Scan, cluster, and understand any unfamiliar repo in minutes | | [