Armory
Source
Browse
Infrastructure

browserbase-bb

Use when an agent must operate the live web — navigate, act, and extract on real pages — via a cloud browser driven by act/extract/observe primitives, with a local-Chromium escape hatch using the same code.

Score
99.4973 signals
Evidence
24,125 stars · 1,664 forks · 3 mentions
Last commit
as last read from GitHub; most reads are from 2 Sep 2026 or later
Listed

Install

No one-command install. Set it up from its source.

Alternatives

  1. open-operator1,953 stars · 325 forks · 1 mention98.469
  2. surf-computer-use856 stars · 141 forks97.631

What it is

A cloud-browser stack: the Stagehand SDK exposes high-level act / extract / observe primitives (do this, pull this data, what's on the page) and runs them against a managed cloud browser. A thin CLI wraps it for shell invocation. It is the canonical way for an agent to operate the web rather than just scrape it.

When to use it

When the task requires interacting with a page — clicking, filling forms, navigating authenticated flows — not just reading static content. For read-only content extraction at scale, a scraping MCP is lighter.

How to install / invoke

Use the Stagehand SDK (or the CLI built on it) and point it at the cloud browser. The same Stagehand code runs against a local Chromium when you want an OSS, no-vendor-lock-in path.

Notes

Browser agents touch live authenticated sessions — the highest-risk surface. Block mutating requests at the network layer, treat extracted page content as untrusted, and require consent before acting on a user's behalf.

Related

InfrastructureStable

e2b-sandbox

Use as the default runtime when an agent must execute untrusted code or commands — Firecracker microVMs with ~150ms cold start give each run an isolated, disposable computer.

13,635 stars · passed install test
Harnessclaudecodex
sandboxruntimefirecrackermicrovm
No one-command install · SourceDetails
MCPsStable

firecrawl-mcp

Use when an agent needs to turn the web into clean data — scrape a page, crawl a site, or extract structured content (incl. from PDFs) as markdown the model can read directly.

7,515 stars · passed install test
Harnessclaudecodexcursor
web-scrapingcrawlingextractioncontent
No one-command install · SourceDetails

Not Indexed

  • browser-agent-security