abeatrix/cline-plugin-jev-browser
User repository on GitHub
Cline Plugin to add a new computer run tool runs by the typesafe/jev model
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Integrations & Adapters
Jev Browser (Cline plugin)
A Cline plugin that gives the agent an isolated Playwright Chromium through a jev_run tool: Jev chooses a concrete operation and target per step via AI SDK 7's experimental_evaluate through the Vercel AI Gateway, and when it picks TYPE_TEXT, gemini-2.5-flash-lite writes the field value.
- Stars
- 0
- Language
- TypeScript
- Reading time
- 3 min
- Published by
- Community
- Author
- abeatrix
- Last verified
- Oct 2026
- Languages
- TypeScript
- Platforms
- PlaywrightBrowserVercel AI Gateway
- Capabilities
- Open sourceRequires API key
Highlights
- jev_run starts a browser, captures the initial screen, runs Jev and returns a final image plus screenshot paths and page states; later calls reuse the browser.
- Jev sees structured DOM observations with indexed visible action targets and selected form options — it does not receive screenshots.
- One evaluation chooses operation and target together, comparing each available action directly against scrolling, waiting and stopping.
- When Jev chooses TYPE_TEXT the plugin calls google/gemini-2.5-flash-lite through the same Gateway key; JEV_TEXT_MODEL can switch it.
- Runs default to 20 steps (maximum 60) with a 100-second cancellation deadline; three non-wait actions without observable progress stop the run.
Watch out
Grounded in the truncated README and GitHub metadata reviewed on 2026-10-15; the excerpt cuts off mid-sentence in the frames and shadow DOM notes, no license is listed in the metadata, and the README warns that both model calls are billable, that an attempted action in a failed run might already have taken effect, and that browser mutations are never retried.
Questions about this entry
Short answers drawn from the linked project or page.
It gives a Cline agent an isolated Playwright Chromium browser through normal Cline tools: you hand it a bounded browser goal with jev_run, and Jev picks the operations step by step until the goal is finished or the run stops.
Through AI SDK 7's experimental_evaluate with typesafe-ai/jev: one evaluation chooses a concrete operation and target together, comparing each available action directly against scrolling, waiting and stopping. Jev sees structured DOM observations, not screenshots.
Jev cannot generate arbitrary text, so when it chooses TYPE_TEXT the plugin calls google/gemini-2.5-flash-lite through the same AI Gateway key to produce the field value. Set JEV_TEXT_MODEL to another Gateway text model ID if you prefer.
Call jev_run with a url, a goal and optionally maxSteps (default 20, maximum 60); it starts the browser automatically and returns a final screenshot plus page states. jev_stop releases the browser and finalizes the video. Further calls reuse the browser until stopped.
AI_GATEWAY_API_KEY in the plugin process environment before starting Cline — keep it in the JSON configuration or process environment, never in browser code. Both the Jev and text-model calls are billable through that one key; no second provider key is needed.
done_unverified, blocked, needs_review, uncertain, step_limit, evaluation_limit or interrupted, and results include steps, elapsedMs and a JSONL tracePath. An attempted action in a failed run might already have taken effect, and browser mutations are not retried, so inspect before continuing.
Runs have a 100-second cancellation deadline, stop after three non-wait actions without observable progress, re-evaluate stale decisions within twice maxSteps, and keep only the last ten actions of entered text in memory. The isolated Chromium deliberately receives no host environment.
More Jev resources like this
Write end-to-end tests for web apps in plain English; jev-e2e turns each case into a plan, has Jev pick the page controls, and runs it with Playwright, returning PASS, FAIL or BLOCKED with HTML, JSON and masked screenshot evidence.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
perixtar/jev-e2eTypeScript
perixtar/jev-e2eTypeScriptA Playwright reporter that uses Jev to score which existing E2E tests a code change could affect and runs only those above a confidence threshold before browser execution, failing open to the full suite.
Jev Integrations & Adapters#community#catalog-import#logicrw
A tool-permissions layer for the Vercel AI SDK backed by Jev through the Vercel AI Gateway: a pre-flight approval gate that decides auto-approve, ask-human or block, plus tool-list pruning that drops irrelevant definitions to save context tokens.
Jev Integrations & Adapters#community#catalog-import#logicrw
Browser QA where Playwright drives and films, and Jev judges: natural-language assertions, self-healing locators and goal-driven click loops, with a JSON-flow CLI for agents, a local dashboard and an agent skill.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
jonymusky/jev-browser-qaTypeScript
jonymusky/jev-browser-qaTypeScriptAn early-alpha integration that lets TypeSafe Jev choose from concrete browser actions while a host - the Codex sidebar or Browser Use's Python engine - handles generation, recovery, and final verification, without patching the installed Browser Use package.
Jev Integrations & Adapters#community#catalog-import#logicrw
A browser and macOS desktop agent for pi: Jev picks each action from structured observations of visible text and addressable controls — never screenshots — inside a bounded loop that stops when stuck or hands control back to a human.
Jev Agents & MCP#community#catalog-import#logicrw
Community
laihenyi/pi-jev-browserTypeScript
laihenyi/pi-jev-browserTypeScriptFrom 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.