Is Browse safe and trustworthy?

tashan scores Browse on public evidence — tashan score 43.0/100, it is currently active, note: a single primary maintainer (bus-factor risk), no known advisories against the current release, it runs a script at install time, its declared dependencies reach: browser. Audit scope: OSV advisories, install-time scripts, build provenance and declared permission surface. We do not review source code or execute the capability.

How do I install Browse?

In Claude Code: run `claude mcp add browse -- npx -y browse-mcp`. In Cursor / Claude Desktop, add it to mcp.json / claude_desktop_config.json.

Who maintains Browse?

source: That1Drifter/browse-mcp.

‹ The Index

Browse

npm

Headless-browser MCP server for any MCP-capable agent (Claude, Codex, Gemini CLI, Cursor, ...). Playwright-based, with accessibility-tree refs, Readability, search, annotated screenshots, and a self-improvement feedback loop. Windows, Linux, and macOS.

Works with: Claude Code, Cursor, Claude Desktop, Codex CLI, Gemini CLI, Cline, Windsurf, VS Code (installable)
installable: each client documents how to load an MCP server of this type — that is the client's promise, not a claim verified against this capability

Work: Agent development

Who it is for: AI engineer

Install (Claude Code):

claude mcp add browse -- npx -y browse-mcp

tashan score: 43.0
Adoption: 37/wk
Upkeep: 58.0
Freshness: 80.0
Evidence coverage: 100% of the inputs this score can use
Health: active

Security audit

Every finding is shown in full — which advisory, the version that fixes it, and the exact command run at install time. Nothing in this audit is behind a licence.

No known advisoriesclearchecked against OSV for 0.8.2

Runs a script at install timecodeplaywright install chromium

Drives a browserfrom declared dependencies

Can carry remote content into your agentit can pull third-party text into the model's context — treat what it returns as untrusted input

Signed build provenancepublished from public CI with an attestation

What changed recently

2026-08-11 browse-mcp now reaches browser Its declared dependencies reach further into your machine than they did before.
2026-08-11 browse-mcp now runs a script when it installs Install-time code executes on your machine before you have used the tool once, and it did not do this before.

You are reading this because you came looking. tashan Pro gives tashan doctor the history behind it, so a run over your own config says which of YOURS moved.

You searched for one. Check the rest of your stack:

npx tashan-cli doctor

Reads the config already on your machine and names what is dead, deprecated or running code at install time. No account, nothing uploaded.

tashan Pro$6/mo

We recorded 2 changes to Browse in the last 45 days. Pro tells you on the day — for the servers in your own config, not the ones you thought to look up.

Every score since we started measuring, for any capability
The named replacement when something you run is dying — not just that it is
tashan doctor over the config you already have, on your machine

Start a 7-day trial › Everything measured on this page stays free.

Show your score

Measured this well? Put the live badge in your README — it updates as the score does.

[![tashan](https://tashan.sh/badge/pkg-browse-mcp.svg)](https://tashan.sh/capability/pkg-browse-mcp.html)

npm ↗ · source ↗ · pkg:browse-mcp

Already running this? npx tashan-cli doctor checks your whole config against the Index — how it works ›

Measured 2026-08-14 · scorer s5 · how · something wrong here?