The Curated Open-Source Jev Ecosystem

174 verified production tools satisfying the Three-Prong empirical benchmark standard.

Showing 174 of 174 curated tools

⭐ Top Verified Utilities Production champions across key decision problem spaces

Ecosystem Leaders

#1 Browser Engine ★ 18,275

browser-use/jev-ultrafast

Evaluates interactive DOM trees via choice in under 100 ms without vision models. 130x cheaper.

Pythonsub-100 ms0 tokens

#1 Context Pruner ★ 6,356

tamaratran/fast-jev-compaction

Scores every tool call result in one fast pass, dropping stale context while keeping active context verbatim.

TypeScript28 ms0 tokens

#1 TUI & Router ★ 114

Dicklesworthstone/skillranker

High-speed terminal triage interface and intent router selecting capable models in single-pass evaluation.

Rustsub-20 ms0 tokens

#1 Autonomous Agent ★ 170

standardagents/jevpilot

Flight navigation and drone trajectory classifier evaluating MuJoCo sensor telemetry in 16 ms cycles.

Python16 ms60 FPS

Complete Curated Directory

This index lists every curated tool grouped by category. The interactive directory above needs JavaScript. Canonical data: jev-ecosystem.json.

Browser & WebMCP Agents (13 tools)

Model & Agent Routing (45 tools)

0xNatoshi/jev-codex-router - Per-turn model and thinking-depth router for Codex. Evaluates user intent via Jev in 28 ms, cutting Codex frontier token consumption by 65%.
AbdelStark/jev-benchmarks - Probability-aware evaluation harness for typed decision models. Measures expected calibration error (ECE) and selective risk.
abhixhek/jevcal - Model calibration and threshold tuning tool. Computes empirical confidence boundaries and monitors distribution drift across production Jev calls.
ably-labs/jev-pong - Real-time Pong benchmark where every ball step is evaluated by Jev vs frontier LLMs through Vercel AI Gateway in 20 ms.
anisselbd/jev-phishing-bench - Adversarial evaluation across 2,000 phishing emails comparing Jev against Claude Haiku 4.5 in 30 ms.
AntonioCoppe/jev-harness - Decision harness for TypeSafe Jev managing policy rules, shadow mode, and confidence gating in 28 ms.
ariel-frischer/jevkit - High-performance Rust CLI and validation library. Evaluates Choice, Score, and Noul questions across 13 offline lint rules before issuing paid API calls, rejecting invalid requests in [sub-5 ms](https://github.com/ariel-frischer/jevkit).
AustinAWay/Working-Memory-Jev - Cognitive working-memory load analyzer for personalized educational content. Uses Jev to evaluate active relational bindings in passages, slashing content validation costs from ~$250,000 to ~$2,000 (125x reduction) vs frontier LLMs.
BillionsBobby/JevRouter - Intent-based orchestration router for Codex, Claude Code, and autonomous agents. Dispatches models, subagents, and tools from a shared candidate set via Jev choice in [25 ms](https://github.com/BillionsBobby/JevRouter) without rewriting agent orchestration.
blakestone-x/jev-mcp - Model Context Protocol server exposing TypeSafe Jev classification, scoring, and matching primitives directly to Cursor and Claude Code in 32 ms.
compozy/yoshi - Context management proxy for Claude Code and Codex. Judges context relevance to user prompts in 30 ms, shrinking prompt size by 60%.
cth9191/jev-compaction-plus - Hardened Claude Code session compaction plugin utilizing Jev keep or drop classifications. Achieves 0.43 to 0.74 s compaction versus 35.1 s for native `/compact` while preserving vital context verbatim.
dbreunig/building-with-jev-skill - Drew Breunig's agent skill for Claude Code and Codex providing prompt patterns and schema templates for System One endpoints in sub-30 ms.
Dicklesworthstone/skillranker - Standalone CLI by Jeffrey Emanuel using TypeSafe Jev to rank agent skills using live session context. Features Claude Code prompt hooks, local feedback, and an interactive TUI running in sub-5 ms, slashing 80%+ of prompt tokens.
gargpratyush/jev-router - Turn-by-turn dynamic model router dispatching between lightweight workers and frontier reasoning models. Evaluates prompt complexity in 28 ms, reducing agent API costs by 78%.
Gaurav-Gosain/jev-sec-bench - Blind security evaluation benchmark testing prompt injection resistance and vulnerability detection in 25 ms (100% on deserialization, 93.5% on SQL injection).
itsmostafa/typesafe-mcp - Zero-dependency Go Model Context Protocol server. Exposes Jev scoring and truth verification to Cursor and Claude in 15 ms.
jev-chat/jev-chat-jarvis - Non-invasive Android and macOS chat copilot. Evaluates user intent, next-action choices, and danger scores over visible conversation screens in sub-50 ms via Accessibility APIs without application tampering.
jkudish/jev-mcp - TypeScript MCP server exposing semantic ranking, claim verification, and screening primitives to agent loops in 30 ms.
kerpopule/hermes-jev-skills - Jev-powered skill suite for Hermes agent frameworks. Implements low-latency model routing, memory pruning, context compaction, and computer-use decision gating.
kitze/skillbox - Self-hosted agent skills library utilizing Jev to recommend tools matching task requirements in 22 ms.
lukstei/slop-grader - Rule-based document grading CLI and agent skill. Uses Jev Score for document-level quality and Choice for line-by-line violation flags, directing agents to remediate style defects in [sub-40 ms](https://github.com/lukstei/slop-grader).
mahlernim/jev-korean-benchmark - Cross-lingual evaluation of Jev on Korean natural language understanding and clinical terminology in 35 ms passes.
manjunathshiva/jev-frontier-bench - Pareto frontier benchmark comparing Jev against 5 frontier LLMs across 200 decisions. Jev scored 72.5% accuracy at $0.025 / 1k calls, while Claude Fable 5.1 cost 478x more ($11.81).
mrnugget/jev-shell-history - Zsh plugin using Jev to select and auto-complete relevant terminal commands from local history in 15 ms.
nanmicoder/jev-arena - Live side-by-side throughput arena comparing Jev against DeepSeek Flash across 10,000 parallel requests in 30 ms median latency.
neural_avb/live-coding - Deep live coding session demonstrating reactive workflows, multi-choice classification outputs, and integrating Jev endpoints into production developer stacks.
qainsights/jev-bench - QA automation and load performance evaluation comparing Jev round-trip latency against generative LLMs for test suite triage.
samtay32/jev-system-architect - System architecture skill for coding agents. Identifies brittle semantic logic in source files and rewrites it into typed choices in 35 ms.
shantanugoel/ask-jev-skill - Agent skill for Hermes allowing autonomous agents to query Jev for bounded decisions in 25 ms.
shitianfang/jev-use - Agent workflow plugin for Claude Code and Codex. Automatically offloads discrete decision steps to Jev in sub-30 ms cycles.
stuartsim/jev-vercel-gateway - Stuart Sim's live latency test routing structured schema payloads through Vercel AI Gateway. Demonstrates sub-100 ms round-trip execution.
supercorp-ai/supercov - Code quality and test coverage analyzer for autonomous coding agents. Employs Jev to score source files in 35 ms to prioritize refactoring.
tamaratran/jev-pruner - Claude Code plugin that uses TypeSafe Jev to filter noisy Bash stdout. Preserves diagnostics and error traces in 25 ms, slashing 70% of low-entropy token spam.
thruwire/foreman - Supervisory coordinator overseeing autonomous Codex runs. Decides whether to continue, verify, or halt execution in 32 ms cycles, saving 90%+ of frontier agent token spend ($0.00004 vs $0.03 per step).
TokenTrim/jev-agent-failure-benchmark - Who&When Pro agent failure attribution benchmark comparing Jev against frontier LLMs on identifying failed steps in 32 ms (130x cheaper).
typesafe-ai/skills - Official collection of agent skills for building with TypeSafe Jev and System One APIs. Includes integration patterns and evaluation templates executing in sub-30 ms.
valentynkit/jev-belay - Pre-completion verification gate for coding agents. Confirms task completion in 25 ms only after tests pass.
valentynkit/jev.nvim - Neovim plugin pairing Treesitter function splitting with Jev to support natural language buffer queries in 18 ms.
vclic/smoking-extraction-benchmark - Clinical extraction benchmark across 1,000 synthetic patient notes. Jev achieved 98.6% accuracy at 1/35th the token cost of OpenAI structured outputs.
vinilana/jev-eval-agent - Public evaluation agent framework for benchmarking System One models across custom task rubrics in sub-40 ms.
wondertwins/jev-benchmark - System One evaluation playground measuring Jev decision accuracy across chess moves and speech-to-text speaker identification in 30 ms.
yusukebe/hono-jev-router - Semantic HTTP router for the Hono web framework created by Yusuke Wada. Matches incoming requests to intent descriptions in 18 ms single passes.
zaidbul/robotic-arm-jev - Real-world robotic arm sorting physical workspace objects in real time. Coordinates evaluated by Jev in sub-50 ms cycles without cloud latency.
zjunlp/JevLoop - Zero-dependency autonomous agent loop from Zhejiang University NLP. Replaces branching LLM prompts with compiled local Jev decisions, eliminating generation latency across [100% of control flow](https://github.com/zjunlp/JevLoop).

Context Compaction & Token Trimming (2 tools)

hemanth/tool-prune - Calibrated tool selection and schema pruning for autonomous agents by Hemanth HM. Filters candidate MCP tools down to relevant subsets using Jev choice and noul in [sub-25 ms](https://github.com/hemanth/tool-prune).
tamaratran/fast-jev-compaction - Experimental context compaction tool evaluated by Nous Research, demonstrating mathematical limits of neural session compaction. *Note:* Heavily disputed across the community due to prompt cache busting and questionable utility in long-running agent workflows (see [Context Compaction Trap](#anti-pattern-1-the-context-compaction-trap)).

Guardrails & Safety Gates (13 tools)

bitnovus/jev-spam-eval - Zero-shot spam and content abuse classifier using Jev boolean questions. Evaluates items in 18 ms.
brainstormity/Jev-Moderation-Bot - Discord moderation bot evaluating message history to detect scam URLs and raid attacks in 35 ms.
david96182/cribrix - Precision-first self-hosted RAG orchestrator using Jev to route queries, filter retrieved passages by relevance, and verify draft claims. Eliminated hallucinations on 22 of 22 unanswerable test queries.
DevMortimer/pi-warden - Security guardrail monitor for Pi coding agents. Intercepts dangerous tool calls and unverified completion claims in 25 ms with zero false-positive rate on safe commands.
EliaAlberti/jev-rules - Dynamic rule injection for Claude Code. Matches edited file paths against architectural guidelines in 20 ms.
eugeniughelbur/jev-engineering - Multi-layer decision policy for coding agents combining deterministic rules with single Jev calls. Runs as a Claude Code hook or MCP server, verified against a [300-call prompt injection test suite](https://github.com/eugeniughelbur/jev-engineering).
Havok (@HavokSocial) - Automated host isolation and firewall containment in sub-50 ms at >= 0.90 confidence with mandatory human sign-off, achieving zero false-positive lockouts.
jomatsu/pi-jev-auto-mode - Auto-mode gate for the Pi coding agent. Evaluates bash commands and file writes semantically in 22 ms to prevent unintended destructive actions.
leepokai/jev-guard - Prompt-injection and dangerous-action detection guardrail for Claude Code and Cursor. Evaluates inputs in 22 ms with 99.4% detection on adversarial jailbreaks.
Pavel Sich (@sichy) - Ambiguity gate in DreamChat intercepting prompts at ≤ 34% confidence in 30 ms to trigger interactive menus, saving $0.05 per confused session.
shiftynick/jev-axi - PreToolUse security gate for Claude Code and Codex scoring destructive command risks before shell execution in 30 ms to stop dangerous execution.
teyhouse/jev-secret-detection - Security benchmark measuring Jev detection accuracy on real API keys and credentials in git diffs in 20 ms.
y0usaf/pi-jev - Pre-execution tool call gate for Pi agents evaluating risk thresholds before file mutations in 28 ms.

Code Review & Git Triage (9 tools)

RAG Filtering & Search Reranking (18 tools)

Real-Time Gaming & Robotics (21 tools)

AboveColin/HA-Jev - Home Assistant integration evaluating home sensor state in 30 ms to trigger household automations.
Bk23544/threejs-grassworks-jev - Real-time AI Scene Director in Three.js by Bilal Khan (@Bk23544). Evaluates freeform natural language prompts ("make it golden hour. Increase the wind") via typed System 1 classification. Dynamically modulates shader uniforms, sun angles, wind gust vectors, and blade sway in under 40 ms at 60 FPS without autoregressive token generation lag.
ChetasLua/jevmeter - Video analysis engine scoring transcripts sentence-by-sentence via Jev in 30 ms. Renders confidence gauges at $0.0015 for an entire 1-hour live stream.
choxos/jevchess - Real-time chess candidate move evaluation without minimax tree search. Predicts candidate blunders and grandmaster moves in under 30 ms per turn.
Diogo Almeida / TypeSafe AI - Real-time 3D combat agent navigating Doom without autoregressive decoding lag. Executes decisions in 35 ms.
fhshaik/typesafe-mario - Emulator control experiment where Jev selects Super Mario Bros actions based on live memory state in 16 ms at 60 FPS.
lbotinelly/jev-little-airways - Island air traffic control simulation where planes observe localized airspace. Jev selects runway approaches and holding patterns in 25 ms.
lukaske/jev-doom-agent - Browser-native Chocolate Doom WASM agent. Evaluates spatial coordinates and enemy vectors in 28 ms to select game actions.
maubaron/typesafe-smash - Real-time 4-player Melee emulation where Jev controls four characters simultaneously from Dolphin RAM at 60 FPS. Total spend was $0.92 across 22 million tokens, compared to $430+ on GPT-4o.
nailthy62/drape-jev - Real-time virtual try-on haul demo for Drape. Matches speech transcripts and outfit metadata in 35 ms at $0.0011 per decision cycle.
phyous/tsai-sc - StarCraft harness evaluating 421 structured tactical decisions in real time in 18 ms.
rmalde/minecraft-agent - Dual-system autonomous agent pairing GPT-6 Astra planning with Jev real-time physical actions. Beats the Ender Dragon in Minecraft Java 1.16.5 in 8 minutes 43.300 seconds across 131 Jev decisions. Cost is under $1.00 total ($0.01 Jev at sub-20 ms latency, $0.96 Astra) with zero deaths.
RomanSlack/jev-drone - Camera-only quadrotor drone simulation on MuJoCo physics. Raycast vectors are scored at 2.5 Hz via Jev choice in 30 ms to avoid dynamic obstacles.
sepiablue-ai/ComfyUI-MiniMax-H3-W4A4-VSA - J-GAS (Jev-Guided Adaptive Sparsity) for MiniMax H3 video generation. Dynamically judges Attention layer importance across 4 Turbo steps and 50 DiT layers (policy `layer_v5`), selecting sparsity retention rates (1%, 3%, 5%, 10%) via Jev choice. Reduces generation time on an RTX 4070 from 6 m 07 s to 3 m 34 s (41.7% speedup, saving 153 s per video). Measured live API spend is $0.0392 for 1,037,063 tokens across 20--30 generated videos (~$0.0015 per video).
socai-io/jev-social - Social media research tool using Jev for typed intent routing and automated sentiment scoring in 25 ms.
standardagents/jevpilot - Autonomous driving simulation selecting velocities and routes in Three.js. Evaluates road hazards in 24 ms cycles.
sunil-sadasivan/jevernetes - Real-time Kubernetes log analysis CLI and local dashboard on [port 8792](https://github.com/sunil-sadasivan/jevernetes). Couples read-only kubectl streams with Jev judgments to flag cluster anomalies and triage pod failure states.
superx-ai/tweet-tester - Production tweet engagement scoring and viral hook predictor powered by TypeSafe Jev. Evaluates draft copy across 30+ viral criteria in a single 42 ms pass at ~$0.04 per 1,000 tweets processed.
TarunTomar122/jev-askable-arm - Simulated Franka robotic arm in ManiSkill physics. Jev evaluates geometric scene state and selects motion primitives in 28 ms.
tdinh_me/tetris-harness - Tony Dinh's empirical reality check on real-time game control. Stripping the TypeScript search harness that pre-filtered candidate placements caused Jev to collapse immediately (lost 0-2, topped out at 0:03). Demonstrates why fast System 1 models require algorithmic search scaffolds and cannot solve open-ended spatial planning alone.
valentynkit/jev-plays-pokemon-red - Game Boy emulator agent where A* code handles pathfinding. Jev makes strategic decisions at gym battles and plot forks in 24 ms.

Finance & Trading Signals (3 tools)

aowang-ai/jev-trade - Hyperliquid automated trading bot querying Jev on each orderbook tick in 25 ms to evaluate sizing, directional choice, and risk.
jarrodwatts/jev-trader - High-frequency automated trading bot evaluating buy/sell decisions on Monad order books in 20 ms (defaults to dry-run mock mode).
phuryn/experiments - Paweł Huryn's 50-edge-case adversarial invoice benchmark across 6 models. Evaluated Jev against GPT-OSS 20 B, Ministral 8 B, and Claude Opus 5, achieving 50/50 accuracy in 30 ms ($0.025 / 1k calls vs $2.83 for Opus 5).

Local Weights & Open Reproductions (40 tools)

benchmarkheaven/jevbench - Benchmark Heaven's independent evaluation suite measuring 42 Jev-class decision systems across 534 decisions (including 220 hard decisions). Ranks systems on the composite JevBench Score (geometric mean of Intelligence, Calibration, Speed, and Cost, 25% each). Jev 1.13.0 leads with a 75.4 composite score ($0.040 / 1k decisions), followed by SemIf Qwen3.5-4 B (74.7 score, ~$0.022 / 1k) and djev Maisa (74.3 score, $0.026 / 1k).
bnsd55/jevmlx - Apple Silicon MLX library generating constrained parallel decisions and schema-valid probabilities from open-weight models in 26 ms.
cloudflare/clef - Cloudflare's open-weight decision model family built on Qwen (Clef 27 B on Qwen 3.8-27B and Clef-flash 9 B on Qwen 3.5-9B). Drop-in compatible with TypeSafe Jev API format. Features multimodal vision encoders for classifying text, JSON, images, and video, trained for 256 k context (64 k hosted). Clef-flash delivers ~39 ms classifications (~13 x speedup over standard Jev) globally across Cloudflare edge GPUs with reinforcement learning (RL) fine-tuning support.
cuth/s1bench - Jake Cuth's empirical benchmark suite evaluating 30+ open-weight Jev alternatives against TypeSafe Jev on DGX Spark. Evaluates decisions/s vs macro accuracy across 1,999 decisions. Shows Jev achieving 77.5% macro accuracy at 2.4 decisions/s (0.42 s/item, ECE 0.076). Reveals Pareto tradeoffs: `simplejev-qwen38-27b` achieves 75.8% accuracy at 1.6 decisions/s, `Reflex-4b` hits 72.0% accuracy at 10 decisions/s, and `Decider-2b` runs 10x faster at 30 decisions/s.
daseinlabs/open-jev - One-pass option scoring on Apple Silicon via MLX. Evaluates choices in 22 ms with interactive Doom demos.
dayhaysoos/jevals - Local evaluation workbench for authoring, running, and diffing Jev benchmark cases across Choice, Noul, and Score rubrics with [69 passing unit tests](https://github.com/dayhaysoos/jevals).
Doccy/Solomon-27B - Open-weight alternative to Jev by [Archer Hume](https://x.com/4rcherhume/status/2101888238357237798) and [Doccy healthcare](https://x.com/4rcherhume/status/2101888238357237798) based on Qwen3.8 [27b](https://x.com/4rcherhume/status/2101888238357237798). Features native multimodal evaluation, a [265k context window](https://x.com/4rcherhume/status/2101888238357237798), multi-choice tagging, and evidence pointer spans.
ekzhang/openjev-sglang - SGLang-compatible prefill-only API implementation designed for ultra-high-throughput parallel choice evaluation on self-hosted GPU clusters in sub-20 ms forward passes.
fastino/GLiNER2.5-Decide - Fastino's 340 M parameter open-weight decision model for constrained classification and joint schema decoding. Evaluates typed rules, mutual exclusions, and character-level entity spans in 38--47 ms on GPUs (167 ms on CPU). Achieved 60.1% accuracy across 17 datasets on the [Fast Decisions Benchmark](https://huggingface.co/datasets/fastino/fast-decisions) (leading 9 of 17 over SemIf, JevK5, and Laya) with 0 generated tokens ($0.00 marginal cost local weights).
featherless-ai/simple-jev - Drop-in server turning open-weights foundation models into non-autoregressive classifiers in a single GPU pass. Available self-hosted or via hosted serverless endpoints at ~$0.002--$0.005 / 1k decisions.
feder-cr/jev - jevos, an open yes/no decision model for laptop CPUs served through llama.cpp behind a Jev-compatible API. Scores text or JSON state against `noul` questions and returns P(yes) with 0 generated tokens. Answers in 50--220 ms on a laptop CPU, including 54 ms against hosted Jev's 344 ms on a short request, at $0.00 marginal cost. Scores 0.815 accuracy against Jev's 0.927 on 2,000 yes/no questions from unseen policies.
FLock.io THIS / THAT Model 1.0 - Binary decision benchmark across [68 questions](https://x.com/flock_io/status/2101936492323209559) recording 94.1% accuracy for FLock's specialized decision model compared to 76.5% for hosted Jev on single-pass classification.
franckverrot/lev - Open reproduction of Jev primitives built on Liquid AI's LFM-2.5 foundation architecture. Powers ultra-compact [350 M parameter](https://huggingface.co/franckverrot/lev-350m) edge deployments for rapid zero-shot choice selection in sub-20 ms.
Heman10x-NGU/Verdict-open-jev - Non-autoregressive decision engine with calibrated uncertainty (RLCD) and in-browser WebGPU playground.
hr98w/jev-visual - Multimodal visual reasoning experiment on Apple Silicon. Evaluates shared image-text contexts with Jev-style constrained choice scoring in 32 ms.
iammrduncan/typesafe-ai-benchmark - Side-by-side benchmark of Jev vs Qwen 3.8 27 B on Cerebras for operational decision questions. Measures 30 ms forward passes and a 95% latency reduction.
ikermoel/open-alternative-jev - Open alternative to Jev emitting typed, calibrated decisions from open-weights LLMs in one forward pass via Hugging Face and vLLM in 26 ms.
jaredpalmer/kev - Open Apache-2.0 decision model family built on Qwen 3.5 (0.8B, 4B, and 9B variants) by Jared Palmer. Replicates Jev's typed contract locally on Apple Silicon and CUDA without cloud API costs; also ported to ONNX and WebGPU by [Nico Martin](https://x.com/nicodotdev/status/2101925774278152664).
kikoncuo/jevfire - JEV-inspired parallel decisions for CUDA LLMs via vLLM API. Evaluates multiple decisions in one context in 18 ms.
lawrence3699/jev-style - Open-weight 0.8B Qwen3.5 decision model serving a `/v1/systemone` API locally via Torch or Apple silicon MLX. Evaluates Choice, Score, and Noul questions in 150 to 200 ms with zero marginal cost.
logan-markewich/jeff - Self-hosted drop-in replacement for TypeSafe Jev powered by a 400 M parameter GLiFormer/GLiNER encoder wrapped in FastAPI. Evaluates local schemas on CPU, Apple Silicon, or cloud GPUs in sub-30 ms.
maisa-ai/djev - Fast structured decision engine powered by Maisa. Evaluates text, JSON, and images into typed yes/no answers, options, and rubric scores. Ranked #1 balanced Jev alternative on S1Bench with slightly faster speed than Jev at an acceptable accuracy tradeoff.
Mapika/decider - Fast decision model reproducing System One using a fine-tuned Qwen3.5-2B base. Emits typed probability distributions across custom categorical choice sets in 28 ms. Achieved 71.0% macro accuracy at 30 decisions/s on S1Bench (10x faster than Jev).
mateolafalce/system-one-model - Open-weights System One architecture and fine-tuning pipeline on Hugging Face.
mizorewww/laya-mlx - Apple Silicon MLX implementation of the Laya non-autoregressive decision engine. Runs offline at 34.8 ms latency and 11.4 decisions/s on M-series chips.
nokia-applied-research/AnyJev - Nokia Applied Research's training-free library that turns open LLMs into Jev-style decision models. Reads typed `choice`, `noul`, and `score` answers from one prefill of the next-token distribution, with nothing generated or parsed. Decides in 7.5--11.0 ms batched and 33.9--49.4 ms unbatched on a 110-token state with Qwen3-8B on one H100 NVL, at $0.00 marginal cost on self-hosted weights. Its calibrated heads score 0.730--0.799 across five Qwen3 models on 2,000 held-out typed decisions, against 0.727 published for Jev and 0.768 for fine-tuned Laya.
olanotolu/jevbetter - Advanced one-pass scorer over variable-length text options using hashed n-gram encoders and rival-aware attention in 20 ms.
ollaya-dev/ollaya - Local runtime that pulls and serves open decision models, including Laya, decider, Kev, NLI, and GLiClass, behind an API wire-identical to TypeSafe's `POST /v1/systemone`. Returns calibrated `choice`, `score`, and `noul` probabilities in one forward pass with 0 output tokens. `laya:en` answers five questions in 8--10 ms on an RTX 4090 at $0.00 marginal cost. The recommended `winnow:e4b` scores 0.722 on typed decisions against Jev's 0.738, in 89 ms for five questions.
OmniJev/PlayJev - 0.8 B parameter distilled fast classifier trained on Qwen3.5-0.8B weights for arcade game control and browser automation. Runs locally on consumer hardware in under 20 ms.
Rizzo-AI-Academy/rizzo-flow - Local decision server that maps each typed question to lettered options and reads only the answer-letter logits from an open model on llama.cpp. Serves `noul`, `choice`, and `score` over a `/v1/systemone` endpoint with 0 generated tokens. Measures 49 / 52 ms p50 / p95 per short decision and 20.99 decisions/s on a shared state on an RTX 5060 Ti at Q8_0. Runs at $0.00 marginal cost, and its fine-tuned 4 B model scores 0.648 against 0.727 for Jev 1.13.0.
taeold/djev-run - Serverless Google Cloud Run deployment for DiffusionGemma-Jev (`djev`). Spins up a Jev API-compatible inference endpoint on an NVIDIA L4 GPU via a single `gcloud run` command in 35--60 ms single-step latency, scaling to $0 when idle.
TheoLeeCJ/SemIf - Open-source research reading choice probabilities directly from open model output heads without text generation. Self-hosted on consumer RTX 3090/4090 GPUs in 18 ms.
TianyuCodings/NanoJev - Open 0.6 B parameter Jev replica based on Qwen3-0.6B weights with parallel decisions and complete probability distributions. Navigates dynamic mazes at 256 decisions/s on Apple Silicon or 4 GB VRAM GPUs.
urchade/GLiNER - Generalist bidirectional transformer encoder framework for zero-shot information extraction and classification without autoregressive LLM decoding. Operates in 15--35 ms on local GPUs with zero generated tokens.
vinnylarouge/jevlike - Training framework optimizing small language models to evaluate variable-length choice sets in parallel. Runs Doom and chess demos in sub-25 ms.
wfzyx/von - Open-source System One decision model running sub-15 ms, non-autoregressive, local drop-in alternative to TypeSafe Jev with RLCD training.
YannQi/Reflex-4B - Distilled 4 B parameter open-weights decision model optimized for rapid single-pass option scoring. Achieves 72.0% macro accuracy at 10 decisions/s on S1Bench, delivering 2--3x higher throughput than Jev.
Yinsongxu/LLM2Jev - Adapts local open-weight language models into Jev-compatible structured decision engines. Exposes Choice, Score, and Noul probability outputs via prefill-only binary inference.
zhangcy122/OpenJev - Open-source alternative to TypeSafe Jev providing typed probabilistic decisions from open-weights models in a single 24 ms forward pass.
zhengxuyu/litjev - Open reproduction of Jev's decision layer on open LLM backends. Exposes a drop-in `/v1/systemone` server with single-pass logit extraction in 25 ms.

Official SDKs & Gateways (23 tools)

AboveColin/jevclient - Asynchronous Python client for TypeSafe Jev with connection pooling and typed schema deserialization.
ainame/swift-typesafe - Unofficial Swift SDK for TypeSafe Jev, supporting iOS and macOS client applications with typed async/await interfaces.
Butochnikov/laravel-typesafe-jev - Laravel 12 and 13 framework integration with typed responses and async request pools.
Cloudflare AI Gateway Integration - Official support for routing and caching TypeSafe Jev requests through Cloudflare AI Gateway.
dannote/jev - Hex package integrating TypeSafe Jev with Elixir OTP. Implements Jev as a GenServer peer for asynchronous pattern-matched message handling.
dfa1/typesafe-java - Modular [JDK 21+](https://github.com/dfa1/typesafe-java) client library for TypeSafe System One endpoints. Includes a dedicated hermetic testkit module for mocking and unit testing application callers.
Gaurav-Gosain/jev-go - Lightweight, dependency-free Go client for TypeSafe Jev with native support for Vercel AI Gateway routing.
gilljon/typesafe-ai-rs - Asynchronous Rust SDK for TypeSafe Jev with connection pooling and typed result decoding.
jamesward/zio-typesafe-ai - Scala 3 and ZIO client library for TypeSafe AI. Provides idiomatic functional programming wrappers for System One endpoints.
joshmn/typesafe-sdk - Community Ruby 3.1+ client library providing typed questions, retry backoff, and model introspection.
kieranklaassen/ruby_llm-typesafe - TypeSafe structured-output provider for RubyLLM 2, enabling typed probability scoring in Rails applications.
mrmps/classifier-dev - Hosted zero-shot classification service and standalone CLI (`npm i -g classifier-dev`) powered by TypeSafe Jev. Features sub-30 ms logit evaluation, multi-input batching up to 1,000 texts, IP rate limiting via Cloudflare Durable Objects, and smart-tier escalation to reasoning models when confidence falls below 0.70.
nshkrdotcom/typesafe_sdk - Idiomatic, type-safe Elixir client for TypeSafe Jev, supporting concurrent request pooling.
obie/ruby_decision_model - Ruby client library providing typed interfaces, retry backoff, and evaluation wrappers for TypeSafe Jev.
OpenRouter Jev Endpoint - Serverless Jev endpoint availability on OpenRouter for unified multi-provider billing.
saibimajdi/typesafeai-dotnet-sdk - Community .NET / C# SDK for the TypeSafe AI System One API supporting typed noul, choice, and score queries.
Shanghua-Gao/RSI-Jev - Open multimodal decision model trained via a recursively self-improving AI research loop. Evaluates typed decisions across text and up to four images on the standard `/v1/systemone` format in a single forward pass.
Stumble/jev-go - Alternative Go client for TypeSafe Jev featuring retry policies and typed schema parsing.
Twister915/typesafe-ai - Typed client for TypeSafe AI in Rust, providing both asynchronous and blocking backends with comprehensive question schema builders.
typesafe-ai/system-one-adapter-python - Official drop-in adapter backed by frontier LLM APIs. Allows developers to simulate and benchmark Jev System One interfaces against traditional chat models.
typesafe-ai/typesafe-sdk-js - Official TypeScript and JavaScript client library (`npm install @typesafe-ai/sdk`) operating in 30--50 ms median latency.
typesafe-ai/typesafe-sdk-python - Official Python client library for TypeSafe AI (`pip install typesafe-sdk`) executing in 30--50 ms.
wiatrM/jevtpp - C++20 library for typed decision routing, classification, and scoring. Evaluates compile-time enum schemas via local Laya in 5.89 ms native FP32 and 10.83 ms ONNX Runtime with 0 generated tokens.

Guides

Video walkthroughs and written guides to Jev from the creators who made them. Each card opens the original.

Interactive Jev ROI & Latency Calculator

Quantify the monthly dollar and compute latency savings of Jev vs frontier conversational LLMs.

Empirical Economics

Monthly Decision Queries 1.0M queries

Range: 100k to 20M queries per month

Average Context Tokens / Query 1,200 tokens

Range: 200 to 8,000 prompt tokens per decision pass

Comparison Baseline Model

Net Monthly Dollar Savings

$5,450 / mo

98.2% Cost Reduction

Monthly Jev Spend

$50.40

Monthly LLM Spend

$5,500.00

Latency Saved / Month

215 hrs

Decision Speedup

32.0x Faster

Interactive Jev Primitive Sandbox

Generate typed code snippets and simulate calibrated probability tensors for choice, score, and noul.

Developer Playground

Context String / State:

Question / Rubric:

Candidate Options (comma-separated):

// Loading...

✓ Decision Evaluated in 21 ms (0 Generated Tokens · Fast Reflex Pass)

Ecosystem Radar: Consensus Candidates

Emerging candidate repositories discovered across 29 independent community mirrors, vetted by automated quality triage.

935 Candidates Tracked

Candidate Repository	Category	Description & Reflex Capability	Community Stars

The First WebMCP-Native Directory

Registered directly on document.modelContext for autonomous browser agents, Claude, and local AI extensions.

View /.well-known/mcp.json →

Domain Calculators & Sizing

Calculates economic ROI, latency differentials, and token savings.

calculate_jev_roi(queries, tokens, model)

Search & Entity Knowledge

Retrieves structured records across 170+ curated tools.

search_jev_tools(query, category, lang) match_project_to_need(need, scope) get_tool_dossier(tool_slug) get_ecosystem_stats()

Form Fulfillment & Actions

Generates verified code snippets and executes sandbox tests.

generate_primitive_code(primitive, ctx, q)

UI Navigation & State

Controls presentation state and category filters.

switch_view_mode(view) filter_tools_by_category(category)

Sub-50 ms Decisions.
$0.042 / M Tokens. Zero Output Tokens.

The Curated Open-Source Jev Ecosystem

browser-use/jev-ultrafast

tamaratran/fast-jev-compaction

Dicklesworthstone/skillranker

standardagents/jevpilot

All Curated Tools (174)

Guides

Interactive Jev ROI & Latency Calculator

Interactive Jev Primitive Sandbox

Reproductions & Interactive Demos Lab

Ecosystem Radar: Consensus Candidates

The First WebMCP-Native Directory

Domain Calculators & Sizing

Search & Entity Knowledge

Form Fulfillment & Actions

UI Navigation & State

Sub-50 ms Decisions.$0.042 / M Tokens. Zero Output Tokens.

The Curated Open-Source Jev Ecosystem

browser-use/jev-ultrafast

tamaratran/fast-jev-compaction

Dicklesworthstone/skillranker

standardagents/jevpilot

All Curated Tools (174)

Guides

Interactive Jev ROI & Latency Calculator

Interactive Jev Primitive Sandbox

Reproductions & Interactive Demos Lab

Ecosystem Radar: Consensus Candidates

The First WebMCP-Native Directory

Domain Calculators & Sizing

Search & Entity Knowledge

Form Fulfillment & Actions

UI Navigation & State

Sub-50 ms Decisions.
$0.042 / M Tokens. Zero Output Tokens.