User repository on GitHub
A small extension that demos the combination of Jev x WebMCP
- Forks
- 9
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Tools & CLIs
Jev × WebMCP Chrome Extension
A Chrome side panel that turns the current page's WebMCP tool schemas into typed Jev questions, predicting the tool call, its arguments, confidence, and latency as you type.
- Stars
- 131
- Language
- JavaScript
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- sdras
- Last verified
- Sep 2026
About this project
A Chrome extension that uses Jev, TypeSafe's System One model, to select and populate WebMCP tool calls from user input. The side panel discovers tools exposed by the current page and displays predicted calls, confidence scores, and latency as the user types. Tool selection is derived from the page's schemas, without site-specific configuration.
Quoted from the project's README. Read the full README.
Highlights
- Tool selection comes from page schemas, not site-specific configuration: tool lists, enums, booleans, and free text each map to a Choice, Noul, or parsed span.
- Confident read-only calls can run automatically; state-changing tools require Enter, and consequential annotations always require a second Enter.
- Jev screens each tool description for instructions aimed at an agent; flagged tools get a badge and cannot run automatically.
- Setup is Chrome 149+ with WebMCP (origin trial or the testing flag), load unpacked, and paste a key from the TypeSafe console; there is no build step.
- Known limits: array arguments populate only the first element, and prediction quality depends on the wording in src/core/questions.js.
Watch out
Apache-2.0-licensed, with the vendored lz-string file under MIT. Needs Chrome 149+ with WebMCP enabled and your own TypeSafe key; host access is granted one site at a time.
Questions about this entry
Short answers drawn from the linked project or page.
Only from your own words. A free-text string parameter becomes a Choice over spans of the user's message, and other numbers come from numbers parsed out of that message in code. Arrays are the weak spot: an array argument such as items populates only its first element for now.
Only for confident read-only tools. Anything without a readOnlyHint annotation is treated as state-changing and needs Enter, and consequential or destructive annotations always require a second Enter regardless of confidence. Automatic execution can be disabled outright, and the second click of a double-click does not count as confirmation.
It can influence selection, which is why the extension screens every tool description for instructions aimed at an agent and badges whatever it flags; flagged tools cannot run automatically. Tool results are displayed in the panel and never included in model requests, and the README is explicit that page-provided descriptions and schemas remain untrusted inputs.
Only where WebMCP exists. You need Chrome 149 or newer with the feature available through the origin trial or behind chrome://flags/#enable-webmcp-testing, and a page that actually exposes tools, or there is nothing to discover. Host access is granted one site at a time, and there is no build step since you load the folder unpacked.
You need Chrome 149 or newer with WebMCP available, either a site in the origin trial or the chrome://flags WebMCP testing flag. Load the folder unpacked from chrome://extensions with developer mode on, open the side panel, paste a key in settings and enable host access one site at a time.
The tool list becomes one Choice over name-and-description plus 'none of these'; enums, consts and oneOf of consts become a Choice over allowed values; a boolean becomes a Noul; an array of enum becomes one Noul per member; free-text strings become a Choice over spans of the user's own words.
Tools are treated as state-changing unless they declare readOnlyHint, so only confident read-only calls can run automatically. Consequential or destructive annotations always require a second Enter, and when tools load Jev screens each description for instructions directed at an agent, badging flagged tools so they cannot run automatically.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Agentic browsing in Chrome
YouTube: AICodeKing tests support routing, refund detection, prompt-injection resistance, and browser automation.
X: Text-to-screenshot (Malay Vasa)
X: Google's MediaPipe Decision Maker, tested in the browser
More Jev resources like this
An experimental Codex skill and JavaScript runtime that has Jev pick macOS Accessibility elements and actions from text candidates, while local policy gates hold sensitive operations for human confirmation.
Jev Tools & CLIs#community#javascript#codex
Community
Sac-Y/Jev-cuJavaScript
Sac-Y/Jev-cuJavaScriptControl a headed Chromium window by voice: one Jev request per partial transcript decides intent, target, and completeness in roughly 300 ms, while code applies thresholds and builds the action.
Jev Tools & CLIs#community#javascript#browser
Unofficial browser automation where the calling LLM states the outcome and Jev decides each element, action, and value from a Playwright page snapshot. Ships as a library, CLI, and MCP server.
Jev Tools & CLIs#community#javascript#browser
Community
Ying-Kai-Liao/jev-browserJavaScript
Ying-Kai-Liao/jev-browserJavaScriptA Chrome and Firefox extension that scores a draft X post with Jev before posting: virality, clarity, ragebait, insult level, regret risk, and typo detection, plus a Send it / Sleep on it verdict.
Jev Tools & CLIs#community#javascript#chrome-extension
Community
RafalWilinski/vibecheckJavaScript
RafalWilinski/vibecheckJavaScriptA DeepSeek Harness plugin where Jev Choice and Noul checks gate every non-read-only tool call - deny confident high-risk calls, escalate ambiguous ones, auto-approve clearly granted reversible ones - and optionally score session messages so valuable context survives instead of being dropped oldest-first.
Jev Tools & CLIs#community#plugin#guardrails
A security hook for coding agents that asks Jev three typed questions before every tool call and scans tool results for prompt injection, with adapters for Claude Code, Codex, Copilot CLI, Gemini CLI, Cursor, pi, OpenCode and ACP.
Jev Tools & CLIs#community#javascript#agents
Community
leepokai/jev-guardJavaScript
leepokai/jev-guardJavaScriptFrom 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.