fuse-browser-usage

Routes mcp__fuse-browser__ tool calls across fetch, crawl, live session, and screenshot profiles.

1|Updated Jul 29, 2026
One-click install
npx skills add https://github.com/fusengine/kimi-code --skill fuse-browser-usage-fusengine
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fuse-browser-usage
Source: https://github.com/fusengine/kimi-code/tree/main/plugins/ai-pilot/skills/fuse-browser-usage
Command: npx skills add https://github.com/fusengine/kimi-code --skill fuse-browser-usage-fusengine

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Agents calling mcp__fuse-browser__ tools often waste time launching full browser sessions to read pages, leak unclosed sessions, loop per-URL calls, or parse snapshots manually. This Skill enforces four zero-tolerance rules and routes every browser call to the correct usage profile before the tool is invoked. ## Core Features & Use Cases - Fast-path routing: Directs reading tasks to browser_fetch, browser_fetch_batch, browser_crawl, and browser_serp_batch without launching a browser, roughly 10x faster. - Session discipline: Enforces one browser_open session, reused sessionId across navigations, and mandatory browser_close on every exit path. - Batch and deterministic extraction: Replaces per-URL loops with batch tools and manual snapshot parsing with browser_extract_schema using containerSelector. - Use Case: After writing code for a login page, an agent opens one session against localhost:5173, asserts zero console errors, checks network responses, fills and submits the form via browser_act, then closes the session. ## Quick Start Before calling any mcp__fuse-browser__ tool, load this skill and pick the matching profile reference for research, webapp testing, or visual capture.

Frequently Asked Questions about fuse-browser-usage

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fetch multiple web pages without opening a browser?▼

Use browser_fetch_batch with a urls array to retrieve many pages in one call, or browser_fetch for a single page. These fast-path tools extract markdown or text without launching a browser session, which is roughly 10x faster.

When should I use browser_open versus browser_fetch?▼

Use browser_fetch, crawl, or serp_batch whenever you only need text, HTML, or structured data. Open a live session with browser_open only when you need interaction, JavaScript rendering, or pixel capture such as screenshots.

How do I test a running webapp in a browser after writing code?▼

Open one session, navigate to your local dev server, assert zero console errors with browser_console, check failed requests via browser_network, interact using browser_act, then always call browser_close. This complements unit and E2E tests rather than replacing them.

How do I take responsive and dark mode screenshots in one call?▼

Call browser_screenshot with a viewports array and a colorScheme parameter to capture multiple breakpoints and dark mode in a single invocation. For several URLs at once, use browser_shots_batch instead of looping single screenshots.

Why is manual browser_snapshot parsing discouraged?▼

Manual snapshot parsing is brittle and non-deterministic. The skill mandates browser_extract_schema with a containerSelector, which extracts correlated fields card-by-card from repeated elements in a deterministic, structured way.

What happens if I forget to close a browser session?▼

Sessions have a 2-minute TTL as a safety net, but the rules require calling browser_close on every exit path including failures and exceptions. Leaking sessions violates the zero-tolerance session discipline.