Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/agentrhq/webcmd/cursorgit clone --depth 1 https://github.com/agentrhq/webcmdWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.01174 |
| Opus 5 | $0.00000 | $0.00587 |
| Sonnet 5 | $0.00000 | $0.00235 |
| Haiku 4.5 | $0.00000 | $0.00117 |
Grade A, and why
cursor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent prompt
Fetch and follow https://raw.githubusercontent.com/agentrhq/webcmd/main/start.md to set up Webcmd end to end.
Manual
Requirements
- Node.js 20.6+
- The
webcmdnpm CLI, installed globally or in the project - A browser runtime; confirm with
webcmd doctorbefore browser work
Install and configure
Install Webcmd and its skills:
npm install -g @agentrhq/webcmd
webcmd doctor
webcmd skills add
When webcmd skills add prompts, choose the agents provider. It installs into ~/.agents/skills/, which Cursor reads on startup along with .cursor/skills/, .agents/skills/, and ~/.cursor/skills/. Cursor then surfaces webcmd-usage and webcmd-browser as skills when a task matches their descriptions.
For a project-scoped setup, copy the webcmd-* skill folders into the project's .cursor/skills/ or .agents/skills/ so the whole team gets them. Restart Cursor after installing skills.
Override default tools
Cursor's agent ships two native web tools:
- Web — generates search queries and performs web searches. This is Cursor's search tool. Keep it on (Settings has a "Web Search Tool" toggle); Webcmd has no search index of its own.
- Browser — navigate, click, type, scroll, screenshot, plus console logs and network traffic. It has dev-server awareness for localhost but can navigate anywhere on the web, so it is the tool that overlaps with Webcmd.
Cursor has no config key that removes the Browser tool for individual users, so steer it with an always-applied rule. Add .cursor/rules/webcmd-browser.mdc:
---
description: Use Webcmd for the open web; keep Cursor's Browser tool for the local dev loop.
globs:
- "**/*"
alwaysApply: true
---
Use Webcmd for anything on the open web — fetching, authenticated
third-party sites, multi-step automation, workflows worth making reusable:
- Check `webcmd list -f json` for an adapter that covers the task; use it first.
- Otherwise run `webcmd --profile work session create "Work Project"`, then drive its returned readable ID with `webcmd --profile work --session work-project-k7 browser tabs` via the shell tool.
- Run `webcmd doctor` first; use `webcmd --profile work session list` to inspect state and `webcmd --profile work session close work-project-k7` when finished.
- For login walls, use Webcmd's human handoff; never type passwords, OTPs, cookies, or credentials.
Use the native Browser tool only for the app being edited: localhost dev server,
console and network triage, visual checks after a change.
Keep using the Web tool to search. Webcmd reads the pages that search finds.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 87 lines · 0 tokens per session scan A ead1a5190080
cursor is an agent published in the GitHub repository agentrhq/webcmd (1,594 stars, last pushed 2d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,174 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
e2e-tester
Tests web applications end-to-end using Glance browser MCP. Navigates pages, fills forms, clicks buttons, takes screenshots, runs assertions, and reports bugs. Use when you want to verify an app works correctly — login flows, forms, navigation, responsiveness — with real browser interaction.
browser-driver
Drives a delegated multi-step browser task through Fast Browser and returns only the distilled result.
test-planner
An expert QA test planner agent that explores web applications and APIs to create comprehensive, risk-prioritized test plans.
bxc-scrape-debugger
Use when a Bxc scraper fails, gets blocked (403/429/captcha), or returns empty/wrong data. Inspects HTML / headers / console dumps and proposes the next profile escalation step.
{{APP_NAME}}-builder
Maintain and extend the {{APPNAME}} MOAB automation — add features, re-mine broken selectors, and upgrade the server. Privileged; use deliberately, not for day-to-day automation.
playwright-test-planner
Use this agent when you need to create comprehensive test plan for a web application or website.