@generativereality/browser-automation 0.6.1 → 0.7.1

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.
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "browser-automation",
3
3
  "description": "Daemonless, per-tab-isolated browser automation for Claude Code via the `browser-automation` CLI. Drives one shared headed Chrome (persistent profile + extensions) over per-target CDP — many parallel sessions, no focus stealing, nothing long-lived to crash.",
4
- "version": "0.6.1",
4
+ "version": "0.7.1",
5
5
  "author": {
6
6
  "name": "generativereality"
7
7
  },
package/README.md CHANGED
@@ -1,12 +1,18 @@
1
1
  # browser-automation
2
2
 
3
- A daemonless, per-tab-isolated browser automation CLI for AI agents (built for
4
- Claude Code, usable anywhere). It drives **one shared headed Chrome** with a
5
- **persistent profile** — cookies and browser extensions (password managers, etc.)
6
- survive across runs — over **per-target Chrome DevTools Protocol**.
3
+ **Let many Claude Code sessions use your own signed-in Chrome at once, each in
4
+ its own tab, without stealing your screen.**
7
5
 
8
- No MCP server, no daemon, **nothing long-lived to crash**. Every command is a
9
- fresh process that connects to a single tab, acts, and exits.
6
+ Your Chrome, your profile: cookies and extensions (password managers, etc.)
7
+ survive across runs, so agents work in the sites you are already signed in to.
8
+ Each session gets its own background tab, so parallel agents never fight over
9
+ one active tab and never pull focus from you.
10
+
11
+ It is a CLI for AI agents (built for Claude Code, usable anywhere) that drives
12
+ **one shared headed Chrome** with a **persistent profile** over
13
+ **per-target Chrome DevTools Protocol**. No MCP server, no daemon,
14
+ **nothing long-lived to crash**. Every command is a fresh process that connects
15
+ to a single tab, acts, and exits.
10
16
 
11
17
  ## Why
12
18
 
@@ -23,8 +29,14 @@ server holding a browser object) has three recurring failures:
23
29
  This CLI fixes all three by talking **per-target CDP** and keeping **no
24
30
  long-lived process**:
25
31
 
26
- - Each command connects to one tab's `webSocketDebuggerUrl` — no whole-browser
27
- enumeration, so target count never wedges it.
32
+ - Each command connects to one tab — no whole-browser enumeration, so target
33
+ count never wedges it.
34
+ - Chrome listens on **no debugging port**. It runs with `--remote-debugging-pipe`
35
+ behind a small broker that `launch` starts, reachable only by your OS user: a
36
+ 0600 socket in a 0700 folder (macOS, Linux), or a named pipe whose random name
37
+ only your user can read (Windows). Login contexts get a broker each. A port
38
+ has no authentication, so any program on the computer could drive a browser
39
+ that listens on one.
28
40
  - Each session owns one tab by name (`-s <name>`); commands only touch that tab.
29
41
  New tabs open in the background and the CLI never activates/foregrounds a tab,
30
42
  so parallel sessions and the user never collide.
@@ -186,7 +198,9 @@ context, set `BAC_CONTEXT` in the env of that project's Claude settings.
186
198
 
187
199
  ## Environment
188
200
 
189
- - `BROWSER_AUTOMATION_PORT` — the debugging port. **Defaults to one per user**,
201
+ - `BROWSER_AUTOMATION_PORT` — the debugging port, used only with
202
+ `BROWSER_AUTOMATION_CDP` and to find a Chrome of yours from before the move to
203
+ the pipe (`launch --restart` switches it over). **Defaults to one per user**,
190
204
  not one per machine: `127.0.0.1` is machine-wide and CDP has no
191
205
  authentication, so a single shared port means the first account to launch owns
192
206
  it and every other account's automation silently drives *that* account's