jev-browser-wingman
MCP wingman that hands an agent's browser clicks and typing to Jev on a real logged-in Chromium, alongside Playwright MCP.
Overview
jev-browser-wingman hands an AI agent's browser clicks and typing to Jev. Jev chooses which element each step acts on and hands the step back when unsure, and wingman carries out the action, while the agent keeps planning and sends its page work to a single tool. It attaches to a real, logged-in Chromium-family browser (Chrome, Edge, Brave and similar) over CDP and sits in front of an existing browser MCP server such as Playwright MCP, hiding that tool's page-action tools while keeping its reading and tab tools.
The README reports 17 small tasks, from ticking a checkbox to checking out in a demo shop, each run twice with wingman and twice with Playwright MCP alone in alternating order, 68 runs in all. Wingman completed all 34 runs versus 32 of 34 for Playwright alone (both misses on a file upload), at $3.75 versus $5.35 total, 30% lower and cheaper on 16 of 17 tasks, but the median run was 0.5 s slower (16.0 s vs 15.5 s). The author notes two runs per task make differences indicative only. It is not affiliated with TypeSafe and has install instructions written for agents.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which element should this step act on? | — |
| Yes / no | Is Jev unsure enough to hand the step back to the agent?(inferred) | yes · no |
How it works
- Agent sends page work to the wingman tool
- Wingman reads the page from the attached browser
- Jev chooses the target element or hands back
- Wingman performs the action
- Agent continues planning
Strengths
- Reported 30% lower cost than Playwright MCP alone
- Completed all 34 runs in the test
- Works with your logged-in browser
Caveats
- Slightly slower median time
- Only two runs per task; indicative results
Getting started
Agents should follow INSTALL-FOR-AGENTS.md in the repository.At a glance
- Type
- MCP server
- Status
- beta
- Language
- TypeScript
- Runtime
- MCP server over CDP
- Has code
- yes
Use cases and integrations
How curated lists describe it
Browser automation where TypeSafe's Jev model picks each step. Attaches to an existing Chrome via CDP and runs alongside Playwright MCP. Not affiliated with TypeSafe.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Which element each browser step acts on, handing back when unsure.
Does it replace Playwright MCP?
It sits in front of it, replacing the page-action tools and keeping reading tools.
Is it faster?
Not on the median (0.5 s slower), but cheaper and completed more runs.
Related projects
MCP 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 Claude Desktop MCP extension that drives Brave or Chrome with Jev choosing on-page actions and a guard freezing irreversible clicks.
MCP serverbetaTypeScriptChrome MCP server with 50 tools that drives your everyday browser without losing logins, where Jev or a local decider picks each step.
MCP serverbetaStealth browser agent where Jev decides each step in about 0.3 s and CloakBrowser carries it out; MCP server, CLI and Python API.
MCP serverbetaPythonLocal MCP bridge letting Codex hand drafting to Gemini via a Google AI Pro quota, with Jev checking drafts against explicit criteria.
MCP serverbeta