cloud browsers run your agent in a datacenter, where logging in is a pain and sites like x and linkedin block you for the datacenter ip
neo runs on your machine instead, so claude code or codex works in tabs signed in as you
free and open source:
github.com/browseros-ai/Brow…
85
claude code's /goal keeps it working until a condition you set is true
/goal all tests in tests/auth pass and lint is clean
after every turn a separate model checks that against the conversation and sends claude back if it isn't met. /goal clear stops it
146
neo isn't a chrome replacement. it's a second browser that sits next to chrome, just for your agents
your logins come over in one click, claude code or codex connects over mcp, and you watch them work live or replay it later
free and open source:
github.com/browseros-ai/Brow…
2
93
claude code can ring your terminal when it's waiting on you, no hook needed
set "preferredNotifChannel": "terminal_bell" in your settings
the default only sends desktop notifications in iterm2, ghostty and kitty, and does nothing in most other terminals
macos 27.3 now takes 14 steps to turn on chatgpt's computer use, up from four
we built browseros neo for the browser side of this: agents drive it over mcp, so the browser part never needs full mac control
As Apple develops its own computer-use agent, it’s making it harder for rival agents to control Macs. In macOS 27.3, enabling ChatGPT’s third-party computer-use support now requires 14 distinct steps, up from four. Apple cites security; Some users see anti-competitiveness.
104
gemini on mac can now take a task across your folders, apps and the web and work it in a live background window, so your screen and cursor stay yours
the catch: it's rolling out in the u.s., in english, to google ai plus, pro and ultra subscribers only
With computer use in the Gemini app for Mac, you can now hand off tasks across your folders, apps, and the web, while your screen and cursor stay yours.
Watch Gemini click, scroll, and type in a live background window, turning a quick sketch into a rigged 3D model in Blender.
117
give claude code a browser in one click
neo finds the agents on your machine. hit connect, restart the agent once, and it gets its own tabs with your logins
free and open source:
github.com/browseros-ai/Brow…
77
put Read(./.env) in claude code's deny list and it can't read the file, not with the Read tool and not with cat
a grep -r across the folder still gets through, the docs say so themselves
deny rules are claude code's check. the sandbox is the os's
117
your agents' sessions, screenshots and history live in ~/.browserclaw/ on your machine
none of it uploads
don't take our word for it, the code is open:
github.com/browseros-ai/Brow…
1
167
neo, the browser we built for your agents, is open source
claude code, codex or any mcp agent gets its own tabs, your chrome logins come over in one click, and every session is saved as a replay you can scrub through
github.com/browseros-ai/Brow…
1
3
334
give claude code a reviewer that can't touch your code
a subagent whose frontmatter says tools: Read, Grep, Glob
claude code enforces that list instead of trusting the prompt, and the subagent runs in its own context, so it judges the diff without the writer's assumptions
1
1
274
claude code's docs have a rule for when it fails the same fix twice
stop correcting. a third correction just adds to a conversation that already holds two failed attempts it keeps re-reading
/clear, then one new prompt that folds in what you learned
1
3
404
make claude code run your typecheck before it can say it's done
a Stop hook runs every time it tries to end its turn, and exit code 2 blocks the stop and sends the compiler output back to it as the reason it can't finish
5
6
572
one click brings your chrome logins, bookmarks and extensions into neo
so your agents start on your real accounts, not a blank sandbox
1
2
250
claude --cloud "fix the flaky test" runs the session in a cloud vm that clones your github remote, so push first
it keeps going with your laptop closed, and claude --teleport pulls it back with its branch and conversation
needs a claude plan login, not an api key
1
276
every session your agent runs in neo is saved on your disk
replay it like a video, with every step it took listed beside it
1
173
neo how-to #8: get claude code to find why your app throws an error
real run in neo on a demo app with a planted bug: it clicked save in its own tab, caught the error the page threw, traced it to line 18 and tested the fix in that tab
2
2
271
start claude code background sessions with claude --bg --name "fix flaky test" "..."
claude agents lists them all on one screen by state, each with its name and current activity
a session lands in ready for review when its pr needs review or has failing checks
2
164











