User repository on GitHub
Fast typed AI decisions on live crypto markets using TypeSafe AI Jev.
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Jev Market Reflex
Bun app that feeds live Kraken BTC, ETH, and SOL order-book features into Jev for typed BUY, SELL, or HOLD calls and runs them through deterministic paper execution with an SSE dashboard.
- Stars
- 5
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- zzsong1023
- Last verified
- Oct 2026
About this project
Fast typed AI decisions on live crypto markets using TypeSafe AI Jev. Jev Market Reflex connects live BTC/USD, ETH/USD, and SOL/USD market data to Jev, turns compact market state into typed BUY / SELL / HOLD decisions, and applies them to a simulated paper portfolio. Real market data. Simulated execution. No real money. Watch the full-quality 16-second demo ↗. Run the recorded demo with no API key and no external services: Open The standard dashboard remains available at.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- Live path reads the Kraken public WebSocket with no exchange account; replay mode runs from a committed JSONL capture with no network.
- Sends three Choice questions in one System One request, then applies confidence gating plus deterministic sizing and exposure caps.
- Paper fills use ask and bid prices with slippage and fees, simulate long, short, and P&L locally, and pause on stale data.
- Keeps the TypeSafe key in the Bun server process and streams decisions to the browser over SSE with state held in memory.
Watch out
Paper-only demo with no wallet, exchange credentials, database, or auth layer; shown P&L is a simulation diagnostic. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
No. The README states displayed P&L is a simulation diagnostic, not evidence Jev predicts markets, and calls the repo an AI systems demo rather than a production trading system or financial advice.
Only live Jev mode needs TYPESAFE_API_KEY. The replay command uses a committed JSONL capture with no key or network, and the Kraken public feed needs no exchange account. The key stays in the Bun server process.
Confidence below 0.55 becomes HOLD, sizing is $100, $250, or $500 by confidence band, exposure is capped per symbol, fills add slippage and fees, and stale market data pauses decisions. Ordinary code owns the rules.
Run bun install then bun run replay and open theater mode. It replays a sanitized capture with preserved timing, while bun test and typecheck cover features, parsing, gating, fills, and accounting.
A single Bun process maintains shallow order books and recent trade flow from the Kraken public WebSocket, computes short-horizon features, sends three Choice questions in one Jev System One request, applies deterministic risk rules, and streams the result to the browser over SSE. Jev returns typed choices, confidence and probabilities; ordinary code owns every execution and risk rule, and all state stays in memory.
Measured on September 19, 2026 with jev-latest and the live Kraken feed: 21 Jev calls, 63 typed decisions, 0 failures and 0 missed cycles, with mean latency 199 ms and P95 332 ms. The README calls it one observed local run rather than a benchmark or latency guarantee, and says results depend on network and API conditions and machine location.
Variables include JEV_MODEL defaulting to jev-latest, DECISION_INTERVAL_MS 500, STARTING_CASH 10000, MAX_POSITION_USD 2000, PAPER_FEE_BPS 5, PAPER_SLIPPAGE_BPS 1, VIDEO_MODE false and REPLAY_SPEED 1. bun run record captures a new bounded real session, 20 seconds by default, into recordings/, which Git ignores; captures hold market features, decisions, probabilities, latency, fills and timestamps, never keys or request headers. VIDEO_MODE optimizes either view for a 1440x900 recording.
More Jev resources like this
3GitHub stars
OpenCode v2 plugin that asks Jev each turn which skills and tools matter for the request, then trims the advertised set for that request only.
Jev Agents & MCP#community#github
Community
universam1/jevselectorTypeScript
universam1/jevselectorTypeScript34.9kGitHub stars
Stencil Labs' coding agent embeds TypeSafe Jev for typed checks, routing, and workflow decisions, with measured Jev tokenizer probes in pi-natives.
Jev Agents & MCP#community#typescript#coding-agent
Community
can1357/oh-my-piTypeScript
can1357/oh-my-piTypeScript1.5kGitHub stars
Open-source platform that puts ACP-compatible agents such as Claude Code, Codex, and Pi inside the team's Slack, Telegram, Discord, and GitHub workflows, with reusable Jev decisions for routing, response gating, and reviewer selection.
Jev Agents & MCP#community#jevbest
488GitHub stars
A coding agent built on pi that hands its non-code decisions — what enters the context, whether a command is safe, whether the work is done — to a small judge model at 35 decision points per turn.
Jev Agents & MCP#community#jevbest
Community
qybaihe/muTypeScript
qybaihe/muTypeScript210GitHub stars
An agent plugin for Pi, Claude Code, Codex CLI and Grok Build that asks Jev two classification questions to decide whether a session sits at a safe compaction boundary, then hints or compacts for you.
Jev Agents & MCP#community#jevbest
572GitHub stars
A CLI plus agent hooks that ask Jev one question per rule in your AGENTS.md on every edit or turn and return a probability, so a violating edit gets repaired inside the same turn. Claude Code, Codex, OpenCode, Pi.
Jev Agents & MCP#community#jevbest
Community
coldteadotai/abideTypeScript
coldteadotai/abideTypeScriptFrom the Jev community
Posts from builders shipping with Jev right now.