docs: add Malkuth listing

README.md

@@ -530,6 +530,7 @@ The [independent model research](community/projects/tools/README.md#independent-
 - [llm-typesafe](https://github.com/simonw/llm-typesafe) - Simon Willison LLM CLI plugin for TypeSafe Jev noul/choice/score (`jev` / `typesafe/jev-latest`). [Project guide](community/projects/tools/llm-typesafe.md).
 - [llmbridge](https://github.com/dragonlin-ai/llmbridge) - OpenAI-compatible LLM routing gateway with TypeSafe Jev as the L2 decision layer. [Project guide](community/projects/tools/llmbridge.md).
 - [macos-computer-use-kit](https://github.com/Sur-Cai/macos-computer-use-kit) - AX-first macOS computer-use MCP/CLI with optional TypeSafe Jev semantic guards before irreversible actions. [Project guide](community/projects/tools/macos-computer-use-kit.md).
+- [Malkuth](https://github.com/newfull5/malkuth) - Open-weight multilingual System One–style decision models (Korean focus) served via Kev `/v1/systemone`. [Project guide](community/projects/tools/malkuth.md).
 - [mayi](https://github.com/AidenHadisi/mayi) - Permission gate for Claude Code/Cursor/Codex: TypeSafe Jev scores tool calls; dialog on unsafe (errors deny). [Project guide](community/projects/tools/mayi.md).
 - [Mechanical Jev](https://github.com/Lasimeri/Mechanical-Jev) - Rust client/`mjev` CLI/eval harness for System One Noul/Choice/Score against local Intel Phi Jev or compatible `/v1/systemone`. [Project guide](community/projects/tools/mechanical-jev.md).
 - [Metis](https://github.com/Ayush0054/metis) - GitHub Action and Python CLI that triage new issues with TypeSafe Jev category labels and missing-detail follow-ups. [Project guide](community/projects/tools/metis.md).

community/projects/tools/README.md

@@ -388,6 +388,7 @@ These projects study related typed-decision patterns using other models. They ar
 | [jev-omni.js](jev-omni-js.md) | Run independent Jev-Omni multimodal decisions in-browser on WebGPU (text+images; WIP video/audio). | JavaScript · onnxruntime-web (Apache-2.0) |
 | [Jevlike](jevlike.md) | Train a small option-attention scorer with synthetic data and optional frozen encoders; independent of official Jev. | Python / PyTorch · research starter |
 | [kevala](kevala.md) | Run Laya/Kev/Bruv/SemIf-style typed decisions in-browser via Rust→WASM + WebGPU (no server; independent of official Jev). | JavaScript/WASM · npm/CDN (`kevala`, Apache-2.0) |
+| [Malkuth](malkuth.md) | Multilingual open decision models (Choice/Noul/Score) via Kev. | Weights · research (Apache-2.0) |
 | [NanoJev](nanojev.md) | Study independent Qwen-based typed decision heads, local serving, and game controllers with recorded comparisons. | Python / PyTorch · model research and replay |
 | [Open Alternative to Jev](open-alternative-jev.md) | Compare packed and separate typed decisions from open models and fit calibration on labeled data; not a Jev reproduction. | Python · Transformers/vLLM research library |
 | [open-jev](open-jev.md) | Experiment with independent Kev and DeBERTa typed decisions locally in a browser; does not use official Jev weights. | TypeScript · npm library, Transformers.js / ONNX |

community/projects/tools/malkuth.md

@@ -0,0 +1,46 @@
+# Malkuth
+
+[All projects](../README.md) · [Independent model research](README.md#independent-model-research)
+
+Open-weight multilingual System One–style decision models (2B/4B, Korean focus): Choice, Noul, and Score over `/v1/systemone` via Kev—adjacent to hosted TypeSafe Jev.
+
+| At a glance | Details |
+| --- | --- |
+| Source | [Source](https://github.com/newfull5/malkuth) |
+| Maintainer | [newfull5](https://github.com/newfull5). Independently curated. |
+| Format | Model cards + serve instructions (weights on Hugging Face `dhtocks/malkuth-*`; served with Kev). |
+| Requirements | Kev (`jaredpalmer/kev`) with uv; HF download for `dhtocks/malkuth-4b` or `-2b`; local GPU/CPU suitable for serve. |
+| License | [Apache-2.0](https://github.com/newfull5/malkuth/blob/af2e1c06ded5c448e78394f356319fc2e49f4c94/LICENSE). |
+| Disclosure | AI-assisted catalog review; no affiliation. Listing is not an endorsement. Live provider paths were not exercised on the review host. |
+
+## When to use
+
+Use to **run a Korean-focused open decision model** compatible with System One question types without the hosted TypeSafe endpoint.
+
+## How it works
+
+Post-trained from Kev; serve with `python -m kev.serve --run dhtocks/malkuth-4b` and POST Choice/Noul/Score questions to `/v1/systemone`.
+
+## Get started
+
+```sh
+git clone https://github.com/jaredpalmer/kev.git kev && cd kev
+uv sync --extra serve
+uv run --extra serve python -m kev.serve --run dhtocks/malkuth-4b --port 8009
+# Model docs/pin: https://github.com/newfull5/malkuth/tree/af2e1c06ded5c448e78394f356319fc2e49f4c94
+```
+
+## Examples and demos
+
+- README curl example for department Choice routing.
+- HF model pages for 2B/4B.
+
+## Limits and data handling
+
+Not TypeSafe-hosted Jev—API-compatible shape via Kev. Eval scores on README are upstream-reported. First load downloads base + fine-tune weights.
+
+## Review and maintenance
+
+Reviewed **2026-09-25** (Europe/Sofia) at [commit af2e1c0](https://github.com/newfull5/malkuth/tree/af2e1c06ded5c448e78394f356319fc2e49f4c94). AI-assisted README and LICENSE inspection; live TypeSafe/provider integration paths not run.
+
+Related: [vLLM Jev](vllm-jev.md), [Mechanical Jev](mechanical-jev.md).