aether-mcp-server 2.0.1 → 2.0.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +23 -60
  2. package/package.json +1 -1
package/README.md CHANGED
@@ -1,28 +1,25 @@
1
1
  # Aether — AI Browser Controller
2
2
 
3
- Aether is an MCP (Model Context Protocol) server that lets AI coding agents inspect and control a real browser through the Chrome DevTools Protocol (CDP). No browser extension needed.
3
+ Give your AI agent a real browser. Aether is an MCP server that lets Claude, Cursor, Kilo Code, Codex, and other AI coding agents inspect and control a live browser through the Chrome DevTools Protocol no extension needed.
4
4
 
5
5
  ## Install
6
6
 
7
7
  ```bash
8
- npx aether-mcp-server
8
+ npx -y aether-mcp-server
9
9
  ```
10
10
 
11
11
  Or globally:
12
12
 
13
13
  ```bash
14
14
  npm install -g aether-mcp-server
15
- aether-mcp-server
16
15
  ```
17
16
 
18
17
  ## MCP Client Config
19
18
 
20
- Add this to your MCP client (Cursor, Claude Code, Codex, KiloCode, etc.):
21
-
22
19
  ```json
23
20
  {
24
21
  "mcpServers": {
25
- "aether-browser": {
22
+ "aether": {
26
23
  "command": "npx",
27
24
  "args": ["-y", "aether-mcp-server"]
28
25
  }
@@ -35,7 +32,7 @@ If installed globally:
35
32
  ```json
36
33
  {
37
34
  "mcpServers": {
38
- "aether-browser": {
35
+ "aether": {
39
36
  "command": "aether-mcp-server",
40
37
  "args": []
41
38
  }
@@ -43,72 +40,52 @@ If installed globally:
43
40
  }
44
41
  ```
45
42
 
46
- That's it. No cloning. No building. No absolute paths.
43
+ No cloning. No building. No absolute paths.
47
44
 
48
45
  ## What It Does
49
46
 
50
47
  - Launches or connects to Chrome, Edge, Brave, or Firefox with remote debugging
51
- - Exposes MCP tools for navigation, clicking, typing, form filling, screenshots, tabs, logs, cookies, network controls, PDF printing, and page inspection
52
- - Uses native CDP input events instead of fragile DOM-only hacks
53
- - Provides Set-of-Marks style visual element references for screenshot-based workflows
54
- - Resolves elements by selector, text, role, accessible name, label, placeholder, XPath, coordinates, same-origin iframe content, and shadow DOM content
55
- - Caches compact page snapshots and invalidates them on DOM/navigation/runtime events
48
+ - Click by text, role, label, or placeholder not fragile CSS selectors
49
+ - Native CDP keyboard and mouse events, not DOM simulation
50
+ - Smart page snapshots that auto-invalidate on DOM changes
51
+ - Set-of-Marks visual element references for screenshot-based workflows
52
+ - Shadow DOM and same-origin iframe support built in
56
53
 
57
54
  ## Browser Setup
58
55
 
59
- Launch a clean browser automatically:
60
-
61
56
  ```
62
57
  launch_browser()
63
- ```
64
-
65
- Pick a specific browser:
66
-
67
- ```
68
58
  launch_browser(browser="chrome")
69
59
  launch_browser(browser="edge")
70
60
  launch_browser(browser="brave")
71
61
  ```
72
62
 
73
- Connect to an existing browser with remote debugging:
63
+ Connect to an existing browser:
74
64
 
75
65
  ```bash
76
66
  chrome --remote-debugging-port=9222
77
67
  ```
78
68
 
79
- Then:
80
-
81
69
  ```
82
70
  connect_browser(mode="connect", port=9222)
83
71
  ```
84
72
 
85
73
  ## Key Tools
86
74
 
87
- - `browser_status` connection and active tab status
88
- - `snapshot_compact` — title, URL, readyState, and interactive element list
89
- - `list_interactive_elements` element refs for click/fill flows
90
- - `click_by_ref` click refs returned by compact snapshots
91
- - `click_text`, `click_role`, `fill_label` semantic actions powered by the locator engine
92
- - `get_state` optional screenshot, tabs, DOM snapshot, and elements
93
- - `get_logs`, `get_network_errors` compact debugging output
94
- - `act` broad compatibility action tool
75
+ | Tool | What it does |
76
+ |---|---|
77
+ | `browser_status` | Connection and active tab status |
78
+ | `snapshot_compact` | Fast title, URL, and interactive element list |
79
+ | `list_interactive_elements` | Element refs for click/fill flows |
80
+ | `click_text`, `click_role`, `fill_label` | Semantic actions |
81
+ | `click_by_ref`, `fill_by_selector` | Direct element targeting |
82
+ | `get_state` | Screenshot, tabs, DOM snapshot |
83
+ | `get_logs`, `get_network_errors` | Live debugging output |
84
+ | `act` | Broad compatibility action tool |
95
85
 
96
86
  ## Project-Local Learning
97
87
 
98
- Aether stores learned lessons and skills in `.aether/` inside your project:
99
-
100
- ```
101
- <project>/.aether/
102
- memory/
103
- lessons.jsonl
104
- learned.json
105
- skills/
106
- <skill-name>/SKILL.md
107
- _registry.json
108
- memory-config.json
109
- ```
110
-
111
- Call `configure_aether_memory` with your project root to enable it. Aether creates `.aether/` and adds it to `.gitignore`.
88
+ Aether stores learned lessons and reusable skills inside your project under `.aether/` lightweight automation notes that make future runs faster. Call `configure_aether_memory` with your project root to enable it.
112
89
 
113
90
  ## Environment Variables
114
91
 
@@ -120,21 +97,7 @@ Call `configure_aether_memory` with your project root to enable it. Aether creat
120
97
  ## Requirements
121
98
 
122
99
  - Node.js >= 18
123
- - Chrome, Edge, Brave, or Firefox installed locally
124
-
125
- ## Architecture
126
-
127
- ```
128
- AI Agent / MCP Client
129
- |
130
- | stdio JSON-RPC
131
- v
132
- Aether MCP Server
133
- |
134
- | Chrome DevTools Protocol
135
- v
136
- Browser Target
137
- ```
100
+ - Chrome, Edge, Brave, or Firefox
138
101
 
139
102
  ## License
140
103
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "aether-mcp-server",
3
- "version": "2.0.1",
3
+ "version": "2.0.2",
4
4
  "description": "Aether MCP Server - AI Browser Controller",
5
5
  "main": "dist/index.js",
6
6
  "bin": {