jev-browser (KesavanKing)
Local browser automation UI where Jev chooses bounded page actions, Browser Harness executes them, and a text model only fills fields.
Overview
Jev Browser by KesavanKing is a local browser automation UI that turns a URL and plain-language goal into safe, bounded actions. It separates three jobs that most browser agents blur: Jev chooses from observed operations only (click, type, select, wait, done or blocked) and their targets; Browser Harness performs the actual Chrome DevTools Protocol actions; and a separate OpenAI-compatible text model supplies field values only when the chosen action is TYPE_TEXT. The model never emits selectors or browser code.
The local web interface shows screenshots with numbered target overlays, ranked targets, probabilities, confidence and the full decision trail, and lets you inspect every decision before execution, switching between manual step-by-step control and automatic runs. Traces can be exported as JSON and screenshots recorded per state, with credentials kept in a local .env. It requires Python 3.12+ and uv.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which operation should be performed next? | click · type · select · wait · done · blocked |
| Choice | Which observed element is the target? | — |
How it works
- Open an isolated browser target
- Observe the page and build bounded candidates
- Jev ranks the next operation and target
- Text model fills fields only for TYPE_TEXT
- Browser Harness executes; UI records the trail
Strengths
- Inspect every decision before it runs
- Clear separation of decision, text and control
- Exportable JSON traces
Caveats
- Requires a separate text-model endpoint for typing
- Local single-user tool
Getting started
`uv sync`, copy .env.example to .env with TYPESAFE_API_KEY and text-model settings, then `uv run jev`.At a glance
- Type
- Project
- Status
- prototype
- Language
- Python
- Runtime
- local web UI with Browser Harness
- Has code
- yes
Use cases and integrations
How curated lists describe it
Local browser automation UI that uses TypeSafe Jev to choose bounded page actions and a text model only for field values.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
The next operation and target from observed page actions.
Does the model write selectors or code?
No, it only picks from observed candidates.
Can I step through manually?
Yes, there is a manual step-by-step mode.
Related projects
RAM-efficient Rust coding-agent harness, listed as using Jev decisions for memory pruning, browser navigation and voice routing.
ServiceproductionRustMCP server and libraries that drive Chrome over CDP and hand a decision model like Jev a bounded menu of page actions.
ProjectproductionTypeScript- Jev Browser★ 335
Headless browser automation via MCP server, CLI or library: Jev picks one action per step and scores goal-met and stuck probabilities.
MCP serverbetaTypeScript Local browser agent that picks each step's operation and element with a Jev-shaped model such as a fine-tuned Laya, no cloud needed.
MCP serverbetaPythonClaude Desktop MCP extension that drives Brave or Chrome with Jev choosing on-page actions and a guard freezing irreversible clicks.
MCP serverbetaTypeScriptBrowser-agent benchmark with five timed events from one prompt, plus a ~200-line DevTools agent where Jev picks each action.
ProjectbetaJavaScript