English · 한국어 · 日本語 · 简体中文

Site: brnyxx.github.io/jev-ra replays a real recorded run and explains the pipeline.

jev-ra

A fast browser-use layer for CLI coding agents. Claude Code, Codex, or any MCP client hands jev-ra a goal. TypeSafe Jev, a System One decision model, picks the operation and the target element for every step in one round trip. Your agent plans, supplies the text values, reads what the page says, and takes over when jev-ra escalates. No second LLM runs inside the loop.

Medians over 5 runs each, 2026-09-18, same machine, same dedicated Chrome, both through OpenRouter. Each run was verified against the final page; 25 of 25 passed with no text-model calls. Method, p90, cost and raw rows.

Quick start

Claude Code

export OPENROUTER_API_KEY=sk-or-...
uvx jev-ra install claude
# then, in Claude Code: "open wikipedia.org and find the Gödel incompleteness article"

Codex

export OPENROUTER_API_KEY=sk-or-...
uvx jev-ra install codex
# then, in Codex: "use jev-ra to open wikipedia.org and find the Gödel incompleteness article"

Shell

export OPENROUTER_API_KEY=sk-or-...
uvx jev-ra doctor
uvx jev-ra run https://en.wikipedia.org/wiki/Main_Page "Open the Godel incompleteness article." \
  --value "search_query=Godel incompleteness theorems"

No Python setup? npx -y jev-ra install claude does the same thing through the npm launcher. The npm package is a launcher only: it finds uv, offers to install it, and runs the PyPI package pinned to its own version.

There is no install step either way: uvx runs jev-ra straight from PyPI and registers uvx jev-ra mcp as the server command. For a permanent copy, uv tool install jev-ra. The key is forwarded from the variable you already exported and is never printed.

No Chrome on the machine either? The repository's Dockerfile builds an image with a Chromium in it: docker build -t jev-ra . then docker run --rm -e OPENROUTER_API_KEY jev-ra doctor. Chrome's own sandbox needs a user namespace a container's seccomp profile usually refuses, so jev-ra launches it once, reads what it said, and starts it again without the sandbox when that is why it would not run.

How it works

  your agent                      jev-ra                              Chrome
 ────────────                ───────────────                      ─────────────
  goal + values  ──────────▶  observe ─────────────────────────▶  snapshot.js
                              │   elements, guards, page marker  ◀──── one eval
                              ▼
                              one request: operation? target?
                              value? prev_ok? goal_achieved?  ──▶  Jev  (~300 ms)
                              │
                              ▼
                              freshness guard ──▶ act ──────────▶  trusted CDP input
                              │                                    (no JS clicks)
                              ▼
                              verify url/title/text/fields
                              │
       Result  ◀──────────────┴── done · blocked · escalate · budget

One decision per step. The only text typed into the page is text you supplied.

MCP tools

Every response carries elapsed_ms, and decisions plus cost whenever Jev was called.

browser_observe returns a page_key; pass it back to browser_click, browser_type or browser_select and a ref from a page that has changed since is refused as stale instead.

CLI

open … close share one browser across invocations through a target id in $XDG_STATE_HOME/jev-ra/session.json. Add --json to any command for the raw payload.

Every run is stored under its own id in $XDG_STATE_HOME/jev-ra/runs/, newest 200 kept. jev-ra trace RUN_ID prints its step table; --html writes a single-file page - the run's own JSON inline, nothing to fetch - to send to whoever asked what the run did.

--profile NAME gives a run a Chrome user-data-dir of its own under $XDG_STATE_HOME/jev-ra/chrome-profiles/NAME, so a login done once on that profile is still there on the next run; open records the profile and the stateful commands reattach to it. The MCP tool takes the same thing as browser_open(url, profile). One process drives one browser, so a server already on a profile refuses a second one instead of answering from the wrong cookies.

serve is the same tool set over MCP's streamable HTTP transport, at /mcp, answering with server-sent events; /healthz answers without a key. Every request carries a key from JEV_RA_SERVE_KEYS as Authorization: Bearer <key>, is charged the decisions it spends against that key's daily quota, and leaves one JSON line on stderr with the run id it also returns in the x-jev-ra-run-id header. A key is never logged: the line names it by a digest.

Python

from jev_ra import Agent

with Agent() as agent:
    result = agent.run(
        "Place the order with express shipping.",
        values={"name": "Ada Lovelace", "email": "ada@example.com"},
        url="https://example.com/checkout",
    )
print(result.status, result.elapsed_ms, [step["target_label"] for step in result.steps])

Values

TYPE_TEXT needs a string, and jev-ra will not invent one. Jev picks which of your values belongs in the field it is about to fill, in the same round trip that picks the field. If nothing fits and no text helper is configured, the run stops with needs_value and reports the field's label, role and current value. You supply the value and call again. The default install has no text model.

When it hands control back

Result.status is done, blocked, escalate or budget. When a run stops short, reason is one of needs_value, stuck_loop, unverified_done, stale, invalid_decision, too_many_controls, provider_error, blocked or blocked_by_site. A run that ends on budget names the budget it hit (steps, decisions or time) in reason instead; provider_error is the provider refusing to answer at all, so check the key and the route rather than retrying the goal. An escalation also carries the top eight operation/target candidates with their probabilities, and up to 3,000 characters of page text — enough to decide what to do without observing again.

A goal that reads as a question — it ends in a question mark, or opens with what, which, how many, when, who or find the — also gets Result.final_answer: one sentence the configured text helper takes from the page the run finished on, whatever the run ended as. With no text helper it stays null and detail says so, and a goal that is an instruction never asks for one.

Verification is deterministic: after every action jev-ra compares url, title, text and field state, and page_changed comes from a semantic page marker, not from the model.

While the page settles after an action, jev-ra asks Jev the next question already, against the page as it should read with that input applied and nothing else changed. If the settled page offers anything the guess did not, the answer is thrown away and the question asked again. Result.speculations and Result.prefetched count how often that paid off.

Benchmarks

Five tasks, five runs each, every run verified against the page it left behind:

Ratios are against browser-use 0.13.10 + gemini-3-flash flash_mode on the same machine and the same Chrome: 23,058 ms, 66,414 ms and 15,071 ms respectively. Text-model calls across all 25 runs: 0. A same-harness re-run of browser-use, five runs per task, was slower still: 9.07×, 8.31× and 7.26×. Our median against browser-use's fastest single run of each task (15,759 ms, 49,914 ms, 17,647 ms) is 5.8×, 5.6× and 4.6×; the headline claim of 3-5× is below that. jev-ra bench --live --runs 5 reproduces this table and prints PASS/FAIL against the v0.1 bar of ≥ 3× on every task with a baseline. Method, the browser-use rows, and how to reproduce them.

jev-ra on the left, browser-use flash_mode on the right, same task, same Chrome, real time:

A run that finishes without doing the task counts as a failure, not as a time.

What it will not do

Each returns an escalation with the page text and the ranked candidates.

FAQ

OpenRouter or a TypeSafe key? Either. jev-ra resolves JEV_RA_API_KEY, then TYPESAFE_API_KEY, then OPENROUTER_API_KEY. A key starting sk-or- selects the OpenRouter route (typesafe/jev-1.13); anything else goes direct (jev-latest). JEV_RA_ENDPOINT and JEV_RA_MODEL override both. OpenRouter is easier to get; direct TypeSafe is roughly 140 ms faster per decision according to the upstream measurements.

What does a task cost? Between $0.00035 (a search, 4 decisions) and $0.00317 (the whole Google Flights flow, 14 decisions). Cost scales with decisions, not with page size, because the state sent is the element table and the visible text, never the HTML.

Does it need its own Chrome? It will find or launch one on its own profile ($XDG_STATE_HOME/jev-ra/chrome-profile) and reuse it; when the Chrome it launched dies, the next command starts another. Point BU_CDP_URL at a different Chrome to override - if nothing answers there, jev-ra says so rather than launching one behind your back. Do not point it at a browser signed into anything you would not let an agent operate.

Why no text model? The host agent already has the context. A second model adds 675-938 ms per field and invents values. You can still configure one with JEV_RA_TEXT_MODEL.

Configuration

$XDG_CONFIG_HOME/jev-ra/config.json sets the same keys; the environment wins.

Credits

jev_ra/browser/snapshot.js and the NEXT_ACTION / TARGET instruction texts are adapted from browser-use/jev-ultrafast (MIT), where they were measured. Chrome is driven through browser-harness (MIT). See THIRD_PARTY_NOTICES.md.

MIT licensed. Contributing · Security · Agent guide · Usage reference · 한국어