How is this different from letting my agent click step by step?

One agent turn instead of six or more: the loop, the re-reading, and the element table stay server-side, and the agent receives the checked result.

Are the speed and cost claims checkable or just marketing?

The documented flight-form goal took seconds of wall time with a handful of typed decisions, and every run prints its own timing line for checking.

What stops the model from inventing a button and clicking it?

The server refuses targets that are not on the page, re-checks actions against the page, and lets a failing code assertion overrule the model account.

What is the catch with free macro replays on changed pages?

Replay makes no model calls and needs no key, but it refuses to proceed when the page no longer matches the recorded path.

How do I install jev-ultrafast-mcp?

Clone and pip install -e . then run scripts/install.py, which finds WorkBuddy, Claude Code, Claude Desktop, Codex, Cursor, VS Code, Cline, Windsurf and Gemini CLI and writes each config, merging and saving a .bak first; it is also on PyPI.

Which jev-ultrafast-mcp tools need a Jev key?

Only browser_goal, the server-side loop, needs a key; browser_open, browser_observe, browser_act, browser_assert and browser_macro need no key, account or network beyond the page itself.

Does jev-ultrafast-mcp bundle a browser driver?

No. It speaks CDP straight to an installed Chromium-family browser with no Playwright, Selenium or screenshot pipeline, and its runtime dependencies are just mcp, websockets and httpx.

CommunityJev Demos & Experiments

Jev Ultrafast MCP

MCP server where one browser_goal call hands a URL, a goal, and code-checked assertions to a server-side loop driven by a decision model over live page elements.

See highlights

Open github.com

Stars: 19
Language: Python
License: MIT
Published by: Community
Author: jiawei686
Last verified: Oct 2026

jiawei686/jev-ultrafast-mcp

User repository on GitHub

View on GitHub

Hand a whole browser task off in one call: a decision model drives the page server-side, so a flow costs one call, not a turn per click. Ref-based element tables, code-checked assertions, zero-model macro replay, over the Chrome DevTools Protocol.

Forks: 2
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

Hand the browser work off — an MCP server that drives the page for your agent. Your agent should not be opening the browser at all. A browser task goes over whole — the URL, the goal, and the check that proves it — and the loop runs on the server. One tool call instead of twenty. Three seconds instead of a minute. A cent instead of a frontier model's context. And it never invents a target: it picks from what the page actually has, and the server refuses rather than guesses.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification Jev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

Server-side loop picks targets from the page element table, never invented selectors.
Code-checked assertions decide success, outranking the model own summary.
Recorded macros replay with zero model calls and refuse on page mismatch.
Speaks directly to an installed Chromium browser with no screenshot or driver stack.

Watch out

Complex or dynamic pages may still need hand-driven tools, and repeated runs depend on the page matching the recorded macro. 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.

More Jev resources like this

open-spark-jev

Independent open-source local decision model inspired by Jev, with a spark-s1 family plus smaller control models for agent gates, served locally with probabilities and policy actions.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

abhishek085/open-spark-jevPython

jev-ra

Browser-use layer for CLI coding agents: Jev picks the operation and target element in one round trip per step, exposed over MCP with a CLI, a Python API and an HTTP server.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

brnyxx/jev-raPython

psearch

A Python search command for terminals and agents: Parallel Search seeds queries, a local Chromium fetches pages concurrently, and Jev scores the evidence to expand a breadth-first link queue. Ships readable output, JSON and an MCP server.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

komikat/psearchPython

ps2-ai-agent

A Python pipeline that captures the PCSX2 window, injects virtual DualShock 2 input and draws a broadcast-style telemetry HUD, with TypeSafe Jev System One as one of three swappable decision engines behind a BaseAgent interface.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

opaielsheikh/ps2-ai-agentPython

open-jev-typed-decision-engine

Open 150M encoder that reproduces the typed noul, choice, and score interface in one forward pass, trainable on a free GPU and runnable locally for nothing.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

intikhab49/open-jev-typed-decision-enginePython

TypeLLM

941

Library that adds schema-guaranteed typed outputs to existing autoregressive LLMs without changing weights, served through SGLang or a hosted API.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

TypeLLM/TypeLLMPython

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.

Jev Demos & Experiments#gaming#agents#control

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai