agent-browser

Name: agent-browser
Author: molefrog

26stars0forks

How to install agent-browser?

molefrog/moi/agent-browser

$npx -y skills add molefrog/moi --skill agent-browser

Installs into the current project.

Summary

Core agent-browser usage guide. Read this before running any agent-browser commands. Covers the snapshot-and-ref workflow, navigating pages, interacting with elements (click, fill, type, select), extracting text and data, taking screenshots, managing tabs, handling forms and auth

Files · 1

View on GitHub

SKILL.md

1	# agent-browser core
2
3	> moi note: vendored from the `agent-browser` npm package (v0.27.0,
4	> `agent-browser skills get core --full` serves the copy matching your
5	> installed CLI). In Claude Code cloud sessions the CLI is preinstalled and
6	> `AGENT_BROWSER_EXECUTABLE_PATH` points at the system Chromium — never run
7	> `agent-browser install` (its Chrome download fails behind the egress relay).
8	> Starting the moi server itself: `docs/browser-testing-cloud.md`.
9
10	Fast browser automation CLI for AI agents. Chrome/Chromium via CDP, no
11	Playwright or Puppeteer dependency. Accessibility-tree snapshots with compact
12	`@eN` refs let agents interact with pages in ~200-400 tokens instead of
13	parsing raw HTML.
14
15	Most normal web tasks (navigate, read, click, fill, extract, screenshot) are
16	covered here. Load a specialized skill when the task falls outside browser
17	web pages — see [When to load another skill](#when-to-load-another-skill).
18
19	## The core loop
20
21	```bash
22	agent-browser open <url> # 1. Open a page
23	agent-browser snapshot -i # 2. See what's on it (interactive elements only)
24	agent-browser click @e3 # 3. Act on refs from the snapshot
25	agent-browser snapshot -i # 4. Re-snapshot after any page change
26	```
27
28	Refs (`@e1`, `@e2`, ...) are assigned fresh on every snapshot. They become
29	stale the moment the page changes — after clicks that navigate, form
30	submits, dynamic re-renders, dialog opens. Always re-snapshot before your
31	next ref interaction.
32
33	## Quickstart
34
35	```bash
36	# Install once
37	npm i -g agent-browser && agent-browser install
38
39	# Take a screenshot of a page
40	agent-browser open https://example.com
41	agent-browser screenshot home.png
42	agent-browser close
43
44	# Search, click a result, and capture it
45	agent-browser open https://duckduckgo.com
46	agent-browser snapshot -i # find the search box ref
47	agent-browser fill @e1 "agent-browser cli"
48	agent-browser press Enter
49	agent-browser wait --load networkidle
50	agent-browser snapshot -i # refs now reflect results
51	agent-browser click @e5 # click a result
52	agent-browser screenshot result.png
53	```
54
55	The browser stays running across commands so these feel like a single
56	session. Use `agent-browser close` (or `close --all`) when you're done.
57
58	## Reading a page
59
60	```bash
61	agent-browser snapshot # full tree (verbose)
62	agent-browser snapshot -i # interactive elements only (preferred)
63	agent-browser snapshot -i -u # include href urls on links
64	agent-browser snapshot -i -c # compact (no empty structural nodes)
65	agent-browser snapshot -i -d 3 # cap depth at 3 levels
66	agent-browser snapshot -s "#main" # scope to a CSS selector
67	agent-browser snapshot -i --json # machine-readable output
68	```
69
70	Snapshot output looks like:
71
72	```
73	Page: Example - Log in
74	URL: https://example.com/login
75
76	@e1 [heading] "Log in"
77	@e2 [form]
78	@e3 [input type="email"] placeholder="Email"
79	@e4 [input type="password"] placeholder="Password"
80	@e5 [button type="submit"] "Continue"
81	@e6 [link] "Forgot password?"
82	```
83
84	For unstructured reading (no refs needed):
85
86	```bash
87	agent-browser get text @e1 # visible text of an element
88	agent-browser get html @e1 # innerHTML
89	agent-browser get attr @e1 href # any attribute
90	agent-browser get value @e1 # input value
91	agent-browser get title # page title
92	agent-browser get url # current URL
93	agent-browser get count ".item" # count matching elements
94	```
95
96	## Interacting
97
98	```bash
99	agent-browser click @e1 # click
100	agent-browser click @e1 --new-tab # open link in new tab instead of navigating
101	agent-browser dblclick @e1 # double-click
102	agent-browser hover @e1 # hover
103	agent-browser focus @e1 # focus (useful before keyboard input)
104	agent-browser fill @e2 "hello" # clear then type
105	agent-browser type @e2 " world" # type without clearing
106	agent-browser press Enter # press a key at current focus
107	agent-browser press Control+a # key combination
108	agent-browser check @e3 # check checkbox
109	agent-browser uncheck @e3 # uncheck
110	agent-browser select @e4 "option-

People who liked this also viewed

6 picks

Type

agent-browser

26stars0forks

View on GitHub

How to install agent-browser?

molefrog/moi/agent-browser

$npx -y skills add molefrog/moi --skill agent-browser

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# agent-browser core
2
3	> moi note: vendored from the `agent-browser` npm package (v0.27.0,
4	> `agent-browser skills get core --full` serves the copy matching your
5	> installed CLI). In Claude Code cloud sessions the CLI is preinstalled and
6	> `AGENT_BROWSER_EXECUTABLE_PATH` points at the system Chromium — never run
7	> `agent-browser install` (its Chrome download fails behind the egress relay).
8	> Starting the moi server itself: `docs/browser-testing-cloud.md`.
9
10	Fast browser automation CLI for AI agents. Chrome/Chromium via CDP, no
11	Playwright or Puppeteer dependency. Accessibility-tree snapshots with compact
12	`@eN` refs let agents interact with pages in ~200-400 tokens instead of
13	parsing raw HTML.
14
15	Most normal web tasks (navigate, read, click, fill, extract, screenshot) are
16	covered here. Load a specialized skill when the task falls outside browser
17	web pages — see [When to load another skill](#when-to-load-another-skill).
18
19	## The core loop
20
21	```bash
22	agent-browser open <url> # 1. Open a page
23	agent-browser snapshot -i # 2. See what's on it (interactive elements only)
24	agent-browser click @e3 # 3. Act on refs from the snapshot
25	agent-browser snapshot -i # 4. Re-snapshot after any page change
26	```
27
28	Refs (`@e1`, `@e2`, ...) are assigned fresh on every snapshot. They become
29	stale the moment the page changes — after clicks that navigate, form
30	submits, dynamic re-renders, dialog opens. Always re-snapshot before your
31	next ref interaction.
32
33	## Quickstart
34
35	```bash
36	# Install once
37	npm i -g agent-browser && agent-browser install
38
39	# Take a screenshot of a page
40	agent-browser open https://example.com
41	agent-browser screenshot home.png
42	agent-browser close
43
44	# Search, click a result, and capture it
45	agent-browser open https://duckduckgo.com
46	agent-browser snapshot -i # find the search box ref
47	agent-browser fill @e1 "agent-browser cli"
48	agent-browser press Enter
49	agent-browser wait --load networkidle
50	agent-browser snapshot -i # refs now reflect results
51	agent-browser click @e5 # click a result
52	agent-browser screenshot result.png
53	```
54
55	The browser stays running across commands so these feel like a single
56	session. Use `agent-browser close` (or `close --all`) when you're done.
57
58	## Reading a page
59
60	```bash
61	agent-browser snapshot # full tree (verbose)
62	agent-browser snapshot -i # interactive elements only (preferred)
63	agent-browser snapshot -i -u # include href urls on links
64	agent-browser snapshot -i -c # compact (no empty structural nodes)
65	agent-browser snapshot -i -d 3 # cap depth at 3 levels
66	agent-browser snapshot -s "#main" # scope to a CSS selector
67	agent-browser snapshot -i --json # machine-readable output
68	```
69
70	Snapshot output looks like:
71
72	```
73	Page: Example - Log in
74	URL: https://example.com/login
75
76	@e1 [heading] "Log in"
77	@e2 [form]
78	@e3 [input type="email"] placeholder="Email"
79	@e4 [input type="password"] placeholder="Password"
80	@e5 [button type="submit"] "Continue"
81	@e6 [link] "Forgot password?"
82	```
83
84	For unstructured reading (no refs needed):
85
86	```bash
87	agent-browser get text @e1 # visible text of an element
88	agent-browser get html @e1 # innerHTML
89	agent-browser get attr @e1 href # any attribute
90	agent-browser get value @e1 # input value
91	agent-browser get title # page title
92	agent-browser get url # current URL
93	agent-browser get count ".item" # count matching elements
94	```
95
96	## Interacting
97
98	```bash
99	agent-browser click @e1 # click
100	agent-browser click @e1 --new-tab # open link in new tab instead of navigating
101	agent-browser dblclick @e1 # double-click
102	agent-browser hover @e1 # hover
103	agent-browser focus @e1 # focus (useful before keyboard input)
104	agent-browser fill @e2 "hello" # clear then type
105	agent-browser type @e2 " world" # type without clearing
106	agent-browser press Enter # press a key at current focus
107	agent-browser press Control+a # key combination
108	agent-browser check @e3 # check checkbox
109	agent-browser uncheck @e3 # uncheck
110	agent-browser select @e4 "option-

People who liked this also viewed

6 picks

Type