Ollaya
Ollama for decision models: pulls and serves Laya, Decider, NLI and GLiClass locally, behind a TypeSafe-compatible API so existing Jev clients work by changing one environment variable.
Open-source apps, extensions, tools and libraries made by the Jev community. Download free ones instantly.
Everything in the store, from apps and extensions to tools and templates, newest first.
Ollama for decision models: pulls and serves Laya, Decider, NLI and GLiClass locally, behind a TypeSafe-compatible API so existing Jev clients work by changing one environment variable.
An infinite canvas that turns LLM conversations into an editable thought graph, now using Jev to pick which past excerpts actually carry forward as context.
A bring-your-own-key AI PR reviewer that runs as a GitHub Action, combining deterministic analysis with model reasoning and a Jev-driven behaviour check before it lets a merge through.
Trading agent that reads Nansen's on-chain whale flow and lets Jev veto a Hyperliquid trade whenever smart money is on the other side of it.
A type-safe Scala client for Jev, added as a new module to sttp-ai, SoftwareMill's existing Scala toolkit for OpenAI, Claude and Gemini.
Collapses a raw log stream into patterns, asks Jev about each pattern once, and routes it to page, ticket, or log using a paging policy you write in plain English.
A short, Jev-driven interview that maps your AI worldview on a doom-to-bloom scale and shows how it compares to public figures like Sam Altman and Elon Musk.
An unofficial, provider-neutral Unix CLI for Jev, written in Go, that pipes state into one question at a time and prints the typed answer as JSON.
An order-flow trading bot where code does the maths and Jev does the judgment, asking 12 typed questions about the market before it agrees to trade.
Hands the "where to click next" decision in Computer Use over to Jev, reading only the interface's text candidates, never a screenshot.
A standalone Rust CLI that puts Jev at the center of skill selection, reading an agent's live context and estimating which of the available skills fits the next step.
A grep that finds lines by what they mean instead of a regular expression, matched by Jev, language-agnostic in both directions.
OpenRouter for agent tools, with a Jev-powered people search that finds and verifies leads for under a cent each.
A live SEO audit tool that judges any website from a single homepage URL using Jev, and exports the findings as PDF, XLSX or Markdown.
An open benchmark suite for Jev-class typed decision models, scoring how smart, cheap, fast, reliable and open each one is.
A Windows desktop chat assistant that reads incoming WeChat messages, judges intent with Jev, and suggests a DeepSeek-written reply.
An open-source desktop browser built around Jev: a language model plans, Jev turns the plan into a structured decision, and Chromium carries it out.
A local MLX model and a hosted Jev call play T-Rex Runner side by side, with live metrics and replay recording, so you can watch the two decision styles compete directly.
Pulls an agent's decisions out of its prose into one typed place, with a confidence score, a code-written gate, a ledger and grading, falling back to Jev only when a rule can't answer.
A crypto market intelligence terminal that reads up to 1,000 tweets about a coin and has Jev turn the market data and sentiment into a buy, sell, hold or take-profit decision card.
A pi agent extension that asks Jev to pick the right reasoning effort for each turn, per prompt and context, instead of running every turn at a fixed effort level.
A pre-execution guard that sits between a terminal coding agent and your shell, asking Jev how likely a command is to destroy something before it runs.
A safety layer that asks Jev five typed questions about a proposed agent action, then lets plain code decide to allow it, block it, or hand it to a human.
A C++20 library that turns application context straight into typed enums, boolean decisions and scores, with a local or remote Jev-style backend behind one API.
Turns any open-weight LLM into a Jev-style decision model, reading typed choices and calibrated probabilities straight from its logits or hidden states.
Connects Claude Code, Claude Desktop, Codex, Hermes and pi to Jev, so an agent gets back a typed judgment with a real probability instead of prose it has to parse.
A fast Jev supervisor that watches your coding agents and decides when a job is really done.
Semantic tool routing, skill discovery and typed Jev decisions for the Pi coding agent.
A second opinion for coding agents in half a second: a Jev command-line tool that can check every command first.
Fewer tokens, more runway: a local proxy that trims stale tool output from long Codex sessions.
A native Swift bridge that lets Swift apps use Jev through Apple's Foundation Models framework.
Agent skills and a coding harness where Jev picks the relevant files, so your coding model only reads what matters.
A code linter that catches logic mistakes by meaning, with rules you write in plain English.
A browser agent that picks its next click from what is really on the page, and backs every answer with a quote.
A large toolkit of agents and skills for coding assistants, with a Jev-powered router that picks the right one for each request.
A decision layer that sits beside your coding agent, shortlisting plans and checking progress in a loop.
Fast, very cheap browser automation: give it a task and a web address, and Jev drives a real browser.
Ten ready-made Jev judgment tools for any agent: verify, screen, find, rerank, classify, decide and more.
An Android copilot for any chat app: it reads the conversation, suggests replies, and you decide whether to send.
A small command-line tool and MCP server for asking Jev typed questions from scripts and agents.
sort, but the key is a description: order lines along any plain-English dimension with Jev.
Browser testing for coding agents without screenshots: Jev picks controls from an index of the page.
Agent evals and guardrails as Jev decisions: one cheap, fast request per trace instead of a language-model judge.
A Go library for Jev, plus a command-line tool for asking it questions inside Unix pipelines.
Ask Jev everyday questions from your code or coding agent with one command, and keep a record of every answer.
Gives Claude Code quick reflexes: Jev scores the possible next steps and hands back the best one.
Puts the unread emails that matter at the top of Gmail and labels them by what they are about.
Lets Jev turn Codex's thinking effort up or down mid-task, so easy steps stop burning tokens.
A web search tool for AI agents that has Jev keep only the results that matter, saving a lot of tokens.
Trims long command output with Jev before Claude Code reads it, so later turns carry less noise.
A local gateway that lets Jev pick which tool your coding agent calls next, and leaves everything else alone.
A browser skill for Codex where Jev does the clicking and Codex does the thinking and checking.
Reliable computer use for any agent, with a Jev mode that drives a desktop app from one plain-language goal.
A fast, cheap second brain for your agent's small decisions, built for Hermes and also usable in Claude Code.
Gives Claude Code, Claude Desktop, Codex and pi direct access to Jev for quick judgment calls.
Did-you-mean suggestions for any command-line tool, matched by meaning rather than spelling.
An early open-source harness that uses Jev to choose context and route work for coding agents.
A plugin that lets you ask Jev yes-or-no, pick-one and score questions from Simon Willison's LLM command-line tool.
A Discord bot that catches spam and scam links as they arrive, judged by Jev instead of keyword filters.
Jev picks which of your standing rules apply to each prompt, so Claude Code only sees the ones that matter.
Browser automation where Claude says what it wants and Jev picks the element, the action and the value.
Use Jev from Elixir as if it were another process: ask from a GenServer, pattern match on the answer.
An open-source Chrome ad blocker with no filter lists: Jev looks at each suspicious element and decides.
Blurs low-effort AI slop on X and LinkedIn as you scroll, with Jev scoring every post.
A simulated drone that flies an obstacle course using only its camera, with Jev deciding what each moment means.
Picks the model and thinking effort for every Codex call, so easy steps stop using your biggest model.
One text box that turns into the right UI as you type: an event card, a checklist, a timer, a poll and more.
A MuJoCo robot decision workbench you run in your own browser, comparing Jev against local and cloud models on real physics tasks with no robot arm required.
Turn your Jev usage into small, very fast classifiers that run locally in Node.js or the browser.
An agent skill that teaches your coding agent when to hand quick judgments to Jev, and where Jev could help your product.
Left-pad a string with Jev. Could it be one line of code? Yes. Does it need a model? No. Anyway.
Say it, and your Mac does it: voice and typed computer use with Jev picking every on-screen action.
Sort documents into categories and split bundled PDFs into their parts, using Jev and rules you write in plain English.
One quick Jev decision on which model, subagent, skill or tool should handle an agent's next step.
A command-line tool that picks the right coding agent, model and effort for each task, then launches it.
Keep the skills list out of every Pi request and let Jev pick the skills each task needs.
Test Jev against another AI judge on real questions, and vote before you see who answered.
Turn documents into knowledge graphs where every relationship is a Jev decision backed by evidence.
One tool that lets an agent find the commands it needs, with Jev searching a tree of capabilities on its behalf.
Learn how Jev makes decisions by running it, in an interactive playground with four real examples.
Run the same batch of comments through Jev and another model side by side, and compare speed, cost and labels.
Watch Jev play chess against language models or Stockfish, or play it yourself, with no key needed.
Identifies which of 261 IRS forms each page of a tax document is, for about a tenth of a cent per page.
A Chrome extension that turns what you type into the right action on any site that supports WebMCP.
Web search in plain language, with Jev choosing where to look and ranking what comes back.
Typed decisions from the command line: jev yes, jev pick and jev score, each with a calibrated probability.
Turn an overloaded Gmail inbox into a small set of useful labels, sorted by meaning with Jev.
Pull the facts a systematic review needs out of research papers, each one as an exact quote you can check.
Sends easy turns to a fast model and hard ones to a strong model in Claude Code and Codex, automatically.
Skips YouTube sponsor reads by reading the captions, with no waiting for someone else to mark them.
A command-line coding helper that learns new workflows over time, with Jev choosing which one to run.
A Claude Code plugin that hides the parts of big tool results you do not need, and brings them back on request.
Fast Jev compaction for the Pi coding agent: prune stale tool history, keep every word of the conversation.
A prose linter that sniffs out AI writing tells, with Jev judging the rules that need reading rather than counting.
A Ruby gem for asking Jev and other decision models typed questions, with no dependencies beyond the standard library.
A chat bar that calls real tools without a language model writing a single word, so it cannot invent a fact.
A smart ad and clutter blocker: Jev spots the junk on a page and the extension clears it away.
Jev finds its way through a Neo4j graph one hop at a time, with an interactive view of the path it took.
Instant context compaction for Claude Code that keeps what matters word for word.
A browser agent that works through a web page in seconds, with Jev choosing every step.
A small skill that lets the Hermes agent, and others like it, ask Jev to settle a decision.
Put a live BS meter on any video: Jev scores every sentence and it is drawn onto a widescreen edit.
Continuous quality scores for code written by AI agents, from a local MCP plugin powered by Jev.
A phone agent that takes one goal and works a real Android app to reach it, with Jev making every decision.
A trading bot that asks Jev to call buy or sell every 300 milliseconds, with a dry-run mode that risks nothing.
A code-review workflow that uses Jev for quick judgments and shows its findings in a calm local dashboard.
Jev controls the original StarCraft shareware with real mouse clicks and hotkeys.
grep for meaning: a command-line tool that answers plain-English yes or no questions about text.