# agentrhq/webcmd

The Browser Infra that learns and cuts token spend by upto 90%

Repository: https://github.com/agentrhq/webcmd
Canonical: https://ross.abutalabs.com/products/webcmd
Homepage: https://webcmd.dev
Language: JavaScript
License: Apache-2.0
License Family: permissive
Topics: browser-automation, cli, ai-agents, browser-use, playwright, self-learning-ai, developer-tools
Last push: 2026-08-26T19:00:26+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 99, longevity 4
- inputs: {"age_days": 62, "days_push": 7, "days_rel": 7, "gap_med": 0, "n_releases_24m": 34}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1579, forks 894 (observed 2026-08-28T04:05:05.623337+00:00)

## What it is
Webcmd is a self-learning browser infrastructure CLI for AI agents that compiles knowledge of websites into deterministic commands, cutting browser-agent token spend by up to 90%. It layers live browser control, sitemap memory, and reusable site adapters so agents stop re-exploring known sites.

## Use cases
- reduce token costs for browser automation agents
- let AI agents control a real browser to click, type, and extract data
- capture a sitemap of a website's pages, actions, and APIs for agents
- author reusable CLI adapters for frequently automated websites
- run deterministic site commands instead of re-navigating with screenshots
- manage browser sessions and cookie profiles for logged-in agent tasks

## When to choose
- you run browser agents that repeatedly visit the same sites and want to cut token spend
- you want deterministic, structured output from web automation instead of fragile screenshot-based reasoning
- you need live browser control plus a path to promote repeated workflows into fast CLI commands
- you use agent harnesses like Codex or Claude and want a browser automation skill

## When to avoid
- you need one-off scraping of unfamiliar sites with no reuse
- you require a mature, battle-tested tool with a long track record
- your automation targets are APIs better served by direct HTTP clients
- you need a GUI-based automation recorder rather than a CLI

## Facets
- artifact type: cli-tool
- maturity: active
- function: agent-framework, cli, web-scraping, developer-tools
- domain: developer-tools, web-development
- platform: cli, cross-platform
- tags: browser-automation, ai-agents, token-optimization, playwright, self-learning, site-adapters, browser-sessions, automation, nodejs

## Member repositories
- agentrhq/webcmd (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:05.623337+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:56:41.559056+00:00, confidence not recorded.
  - readme: https://github.com/agentrhq/webcmd (fetched 2026-08-28T04:05:05.623337+00:00, sha 64a61af8334a)
  - homepage: https://webcmd.dev (fetched 2026-08-29T11:27:11.047194+00:00, sha fea0116bca9f)
  - site_page: https://webcmd.dev/docs (fetched 2026-08-29T11:27:11.056498+00:00, sha eae7e31f59b1)
- Data as of 2026-08-30T08:39:29.467469+00:00.