User repository on GitHub
A tool calling chat bot built with Jev and no LLM.
- Forks
- 24
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Cookbooks
Jev Chat — w3cj
A chat-shaped command bar where Jev picks the tool, arguments, confirmation, and reply type each turn, and code calls the MCP servers, so no LLM ever writes the reply text.
- Stars
- 117
- Language
- TypeScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- w3cj
- Last verified
- Sep 2026
About this project
A chat-shaped command bar that calls real tools, without an LLM writing anything. Every turn, a classifier picks: what was asked, which tool to call, which value goes in each argument, whether to confirm first, and what kind of reply to give. Code does the rest: it calls the MCP servers and builds the reply from the tools' own data. Because no model ever writes the text, the assistant cannot invent a fact: every value on screen was either typed by the user or returned by a tool.
Quoted from the project's README. Read the full README.
Works forJev for Tool calling
Used inJev for Data & Analytics
Highlights
- Because no model writes the text, the assistant cannot invent a fact: every value comes from the user's message or a tool result.
- An inspector pane shows the request, every question, every probability, and what the code did with the answers.
- Built-in MCP servers cover weather, units and maths, Wikipedia, and recipes; Brave Search, Todoist, and Home Assistant need their own keys.
- Needs Node 24+ and pnpm 11 and stores conversations in SQLite; only TYPESAFE_API_KEY is required for the app to work.
- The API listens on 127.0.0.1 only and rejects requests from other hosts; the app works in English only.
Quickstart
bash
pnpm install
cp .env.example .env
pnpm devWatch out
MIT-licensed code, while fetched data keeps its own terms (Open-Meteo CC BY 4.0, Wikipedia CC BY-SA 4.0). Needs Node 24+, pnpm 11, a TypeSafe key, and a C++ toolchain if better-sqlite3 has no prebuilt binary.
Questions about this entry
Short answers drawn from the linked project or page.
Only as far as your own words and tool data reach. Every text argument is copied from your message, the conversation, or an earlier tool result, so a request that cannot be mapped to a known tool gets a what-I-can-do reply rather than an invention. It also works in English only.
No. With no login, the API listens on 127.0.0.1 only and refuses requests addressed to, or sent from a page on, any host other than localhost, 127.0.0.1, or [::1]. It is also a dev setup: pnpm build builds only the web app and nothing serves those built files.
Nothing else. Weather, units and maths, Wikipedia, and recipes ship as bundled MCP servers that need no credentials, and each still works without Jev. Brave Search, Todoist, and Home Assistant show as having no key in the sidebar and stay unavailable until you add their own, and without a Jev key every message simply gets an error reply.
Possibly. better-sqlite3 is a native module, and if no prebuilt binary matches your platform pnpm install compiles it and needs a C++ toolchain. On top of that you need Node 24 or later and pnpm 11, which corepack enable picks up from the version pinned in package.json.
Use Node 24 or later and pnpm 11, which corepack enable picks up from the version pinned in package.json. Then pnpm install, cp .env.example .env, and pnpm dev, which serves the API on :8787 and the web app at http://localhost:5173.
It is plain code reading the answers in order: cancel or confirm a pending action, small talk, which tool, a missing argument, a confirmation, then the call. When confidence is below the CONFIDENT bar and the runner-up is within CLOSE_MARGIN, it offers two buttons instead of choosing.
A typed message goes through two rewrites, spell check then follow-up resolution. Both do what they can in code and only ask Jev when there is more than one reading; neither runs for a button click or without a Jev key, and if either Jev request fails the step is skipped.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev in front of the HeyGen MCP
X: Playwright CLI + Jev vs Playwright MCP
X: A local MCP server that cuts computer use from 23s to 0.8s. Cartwright built Cortex, a local MCP server where…
X: A side panel that drives any site's WebMCP tools
Other listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
Research decoder that treats Jev as if it were a language model, assembling replies in code from Choice distributions over a hierarchical codebook instead of generated tokens.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
adhyaay-karnwal/jev-chatPython
adhyaay-karnwal/jev-chatPythonWindows and Android chat app for Jev via Vercel AI Gateway that answers document questions with per-option probabilities and costs.
Jev Demos & Experiments#community#github
Community
Manta-Boardgame/jev-chatHTML
Manta-Boardgame/jev-chatHTMLMore Jev resources like this
A self-hosted AI trading platform that can put Jev in front of strategy and Quick Trade entry orders as an auditable PASS/REJECT gate, with LLM fallback and deterministic bypasses for exits and protective orders.
Jev Cookbooks#community#python#trading
Community
OpenByteInc/QuantDingerPython
OpenByteInc/QuantDingerPythonDescribe a startup idea and Jev returns KILL IT, FIX IT, or SHIP IT from one request of ten scored questions, with goal-weighted scoring and a SQLite archive.
Jev Cookbooks#community#typescript#demo
Community
monteduro/killmyideaTypeScript
monteduro/killmyideaTypeScriptAn SO-101 workbench where an evaluation model or rules baseline picks bounded joint steps from typed candidate actions: decision runs are spend-capped via the Vercel AI Gateway and drive a real LeRobot arm or mock hardware.
Jev Cookbooks#community#typescript#robotics
Community
grmkris/robo-harnessTypeScript
grmkris/robo-harnessTypeScriptA hands-on lab that arranges thirty real uses of Jev into ten levels: levels 1–5 call Jev from plain code, and levels 6–10 run inside the pi coding agent, ending with an agent that writes its own Jev questions.
Jev Cookbooks#community#github#typescript
Community
disler/ten-levels-of-jevTypeScript
disler/ten-levels-of-jevTypeScriptA self-hosted, versioned skills library for AI agents with HTTP MCP, scoped client keys, and opt-in Jev recommendations that rank skills for a task using your own OpenRouter, TypeSafe, or Vercel key.
Jev Tools & CLIs#community#typescript#mcp
Community
kitze/skillboxTypeScript
kitze/skillboxTypeScriptA local-first MCP server that gives Claude Code, Codex, Cursor, and OpenCode structured quality scores from Jev across correctness, complexity, tests, security, and other dimensions while the agent writes code.
Jev Tools & CLIs#community#typescript#mcp
Community
NiazMorshed2007/jev-reviewTypeScript
NiazMorshed2007/jev-reviewTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.