Let agents use a browser
You end with an agent that checks a local page in a tab beside its thread, which you can watch, take over and screenshot.
- SpecWithCode
- 1:10
- Codex
- Narrated, with captions
Recorded on SpecWithCode 1.0.47, 18 October 2026, added 18 October 2026. The current release is SpecWithCode 1.0.44.
Before you start: SpecWithCode for Mac; a signed-in coding agent (Codex in this video); a project with a web page served locally.
Steps
- Open Settings, Integrations and turn on Agent browser access. Projects can override it, and the change applies when an agent session next starts.
- Serve your page locally (greet-app's demo page runs on port 5173 in this video).
- Ask the coordinator to check the page in the browser beside the thread, and say you want to watch; it starts a worker thread for the check.
- Open the thread. Its agent opens an isolated tab with the
preview_*tools; dock the floating tab to keep it beside the thread while the agent types and clicks. - Click Take over to drive the same tab yourself while the agent waits, then Hand back.
- Ask for a screenshot: the agent captures the tab as you left it and shows it in the chat.
Transcript
Built from the caption file of the narration.
0:00 What you will see
Let agents use a browser, right beside the thread.
0:04 Turn on Agent browser access
Open Settings, Integrations, and turn on Agent browser access. Projects can override it, and it applies when an agent session next starts.
0:14 Ask for a browser check
greet-app serves its demo page locally, on port 5173. Ask the coordinator to check it in the browser beside the thread, so you can watch.
0:26 Watch the agent drive
Open the new thread. Its agent opens an isolated tab with the preview_* tools. Dock the tab beside the thread and watch it type Ada and press Greet. The page says Hello, Ada.
0:42 Take over the tab
To try something yourself, click Take over. It's the same tab, and the agent waits while you drive. Type Grace, press Enter, then Hand back.
0:53 A screenshot in the chat
Now ask for a screenshot. The agent captures the tab as you left it and shows it in the chat.
1:03 Next: your real Chrome
To use your real Chrome instead, see the SpecWithCode Browser extension docs.
Learning path
Run SpecWithCode
Understand the setup, run a project through its coordinator, then add plans, memory and a second provider.
See how SpecWithCode runs your agents
2:54, SpecWithCode
Start SpecWithCode with a coordinator and two workers
1:48, SpecWithCode
Run a project with a coordinator and workers
1:29, SpecWithCode
Plan, verify and schedule project work
1:35, SpecWithCode
Keep project memory and read honest usage
1:06, SpecWithCode
Continue a thread on Codex at a usage limit
1:01, SpecWithCode
Keep working when a provider runs out
1:39, SpecWithCode
1:13, SpecWithCode
Finish a thread with a pull request and auto-merge
1:19, SpecWithCode
Let agents use a browser (this guide)
1:10, SpecWithCode
One more guide is planned for this path.