npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@graphmemory/browser-client

v0.1.0

Published

Tiny dependency-free client for the browser-mcp stateless REST API (/api/v1).

Readme

@graphmemory/browser-client

A tiny JS/TS client for the browser-mcp stateless REST API (/api/v1) — zero runtime dependencies (uses the global fetch). Drive the same live browser an MCP agent uses, from a plain Node script, and get structured JSON back without the MCP handshake.

Node 18+ (uses the global fetch). Targeted at server-side scripts; for browser-tab usage the server needs a matching --cors-origin and CORS preflight support (not provided by default).

Install

npm install @graphmemory/browser-client

Usage

import { BrowserClient } from "@graphmemory/browser-client";

const b = new BrowserClient({
  baseUrl: "http://127.0.0.1:7777",
  apiKey: process.env.BROWSER_MCP_API_KEY, // only if the server requires auth
  profile: "work",                          // the shared browser to drive (default: "default")
});

// Ergonomic shortcuts for common tools:
const opened = await b.open("https://example.com");
console.log(opened.data);                   // structured TabInfo: { tab_id, url, title, status }

const page = await b.read({ mode: "markdown" });
console.log(page.data?.content);            // the article as Markdown (structured, not scraped text)

// Any tool by name (full surface) — args match the MCP tool schema:
const log = await b.tool("browser_network_log", { min_status: 400 });
console.log(log.data);                      // { entries: [...], total }

Working alongside an agent

The client is stateless: each call is an independent POST. profile selects which shared browser to drive — use the same profile name as the agent and you share its cookies, tabs, and network log. Tabs are shared and addressable; pass an explicit tab_id to act on a specific one (your "active tab" is independent of the agent's):

type TabsData = { tabs: Array<{ tab_id: string; title: string; url: string }>; active: string | null };
const tabs = await b.tabsList();
if (tabs.ok) {
  const tabId = (tabs.data as TabsData).tabs[0].tab_id;   // response `data` is typed as `unknown`
  await b.read({ tab_id: tabId });                         //   — per-tool output schemas aren't in OpenAPI v1
}

Error handling

  • Tool-level outcomes (a failed browser_expect, an element not found, …) resolve normally with { ok: false, error } — HTTP is still 200.
  • Transport errors (bad args → 400, auth → 401, unknown tool → 404, cap → 503) throw a BrowserClientError carrying .status and .body.
import { BrowserClientError } from "@graphmemory/browser-client";

const r = await b.tool("browser_expect", { assertion: "visible", target: "#x" });
if (!r.ok) console.warn("assertion failed:", r.error.message);

try {
  await b.open("not-a-url");
} catch (e) {
  if (e instanceof BrowserClientError) console.error(e.status, e.body);
}

API

  • new BrowserClient({ baseUrl, apiKey?, profile?, fetch? })
  • tool<Name>(name, args?, { profile? })ToolEnvelope — call any tool. Both the tool name and args are type-checked against the OpenAPI spec (see Types below).
  • Shortcuts: open(url, opts?), read(opts?), click(target, opts?), type(target, text, opts?), snapshot(opts?), evaluate(expression, opts?), tabsList(opts?). Each shortcut's opts type is derived from the same schema (e.g. open accepts Omit<ToolArgs<"browser_open">, "url">).
  • listTools(){ tools: [{ name, description }] }
  • openapi() → the OpenAPI 3.1 document.
  • releaseProfile(profile?) → release the profile's REST holder. The browser is shut down only if no MCP session still holds it (ref-count); otherwise just the REST holder is dropped.

For the error model see Error handling above.

Tool catalogue

All 36 tools are reachable through b.tool(name, args). Authoritative sources for parameter shapes:

Navigation | Tool | Purpose | |---|---| | browser_open | Open a URL in a new tab (or navigate an existing tab). | | browser_back / browser_forward / browser_reload | History navigation. | | browser_open_visible | Open a URL in a visible window for manual interaction (sign-in, CAPTCHA). | | browser_scroll | Scroll the current tab. |

Reading content | Tool | Purpose | |---|---| | browser_read | Page as Markdown (default), text, or HTML; optional selector to scope. | | browser_snapshot | Accessibility tree (compact YAML/JSON; supports diffing via store_as/diff_against). | | browser_find | Search visible text; returns snippets + best-effort selectors. | | browser_evaluate | Run a JS expression in the page; returns the JSON-serialized result. |

Tabs | Tool | Purpose | |---|---| | browser_tabs_list | List open tabs. | | browser_tab_switch | Make a tab active. | | browser_tab_close | Close a tab. |

Interacting | Tool | Purpose | |---|---| | browser_click | Click an element (locators: text/role/label/placeholder/testid/selector). | | browser_type | Fill an input/textarea/contenteditable. | | browser_press | Press a key or chord (Enter, Control+A, …). | | browser_hover | Hover an element to reveal menus/tooltips. | | browser_select_option | Select option(s) in a native <select> by value/label/index. | | browser_check | Set a checkbox/radio (idempotent — unlike click). | | browser_drag | HTML5 drag-and-drop. | | browser_fill_form | Batch-fill multiple fields and optionally submit. |

Waiting & asserting | Tool | Purpose | |---|---| | browser_wait | Wait for a selector state or a JS condition. | | browser_expect | Assert a condition (visible/text/value/url/title/count); retries until timeout. |

Capture & files | Tool | Purpose | |---|---| | browser_screenshot | PNG screenshot (viewport or full page); base64 in the content block. | | browser_save | Save the page as PDF / MHTML / HTML to the server's filesystem. | | browser_upload | Upload file(s) to an <input type="file"> (paths on the server). | | browser_download_wait | Trigger a download and capture the resulting file. |

Browser state (per profile) | Tool | Purpose | |---|---| | browser_cookies | get / set / clear cookies in the profile. | | browser_storage | localStorage / sessionStorage get / set / remove / clear. |

Browser config & permissions | Tool | Purpose | |---|---| | browser_configure | Viewport, UA, locale, color scheme, mobile preset, extra HTTP headers, … | | browser_permissions | Grant camera / microphone / geolocation / clipboard / notifications. | | browser_set_geolocation | Set the emulated coordinates (pair with browser_permissions). | | browser_handle_dialog | Policy for the next native alert / confirm / prompt. |

Network & console inspection | Tool | Purpose | |---|---| | browser_network_log | Ring buffer of recent requests (filter by tab / URL regex / status / method). | | browser_console_log | Ring buffer of console.log/info/warn/error/debug + pageerror. | | browser_network_body | Captured response bodies (small JSON/text only, last 50). |

Tools that operate on the server's filesystem (browser_save, browser_upload, browser_download_wait) read/write paths on the host running the daemon, not the script's machine — sandboxed by the same IO guards as the MCP surface.

Types

Argument types are generated from the server's OpenAPI spec by openapi-typescript — committed under src/generated/openapi.ts. The package exports:

  • ToolName — the union of all 36 tool names.
  • ToolArgs<Name> — the argument shape for a given tool, taken straight from the spec.
import type { ToolName, ToolArgs } from "@graphmemory/browser-client";
type OpenArgs = ToolArgs<"browser_open">;  // { url: string; tab_id?: string }

Misspelled tool names and bad arguments fail at compile time:

await b.tool("brwoser_open", { url: "..." });
//           ~~~~~~~~~~~~~ ✗ TS2322: not assignable to ToolName

await b.tool("browser_select_option", { target: "Country" });
//                                    ~~~~~~~~~~~~~~~~~~~~ ✗ TS2741: missing 'values'

await b.tool("browser_open", { url: "https://example.com", tabid: "t1" });
//                                                         ~~~~~ ✗ TS2353: did you mean 'tab_id'?

To regenerate after the server adds/changes a tool:

cd apps/browser-mcp && npm run openapi              # emits apps/browser-mcp/openapi.json
cd ../../packages/browser-client-js && npm run gen  # regenerates src/generated/openapi.ts

License

Elastic-2.0.