Record and Replay Workflows
Record your browser actions once, replay them forever. The recorder captures what you do and creates replayable workflows with self-healing selectors.
The recorder is functional but still being refined. Some complex workflows may require manual adjustment. Feedback helps us improve it.
How logins persist
Browser Agent uses a persistent browser profile stored at:
~/.rird/browser-profile/ When you log into any site (Gmail, LinkedIn, etc.), that login persists across all future sessions. You don't need to re-login each time - the browser remembers your sessions just like Chrome does.
First time: Run any task that requires login. Browser opens, you log in manually when prompted.
After that: The login is saved. Future tasks (and recordings) use the same session automatically.
Multiple sites: Each site's login is stored independently. Log into LinkedIn once, Gmail once, etc.
How recording works
Run the record command with a description of your goal.
A browser opens (already logged in if you've logged in before). Click, type, navigate - do the task yourself.
Press Ctrl+C when done. The recording is saved with a unique ID.
Run the replay command with the recording ID. Uses the same persistent login session.
Commands
rird record "fill out the contact form on example.com" Opens browser and starts capturing your actions.
rird replay "abc123" Replays the workflow with the given recording ID.
rird --replay "abc123" --replay-input data.csv Replays the workflow multiple times with different input values from CSV.
rird --record "submit expenses in SAP" Flag-based syntax works too.
Where recordings are stored
~/.rird/recordings/Each recording is saved as a JSON file with the recording ID as the filename. You can view, edit, or delete recordings directly from this folder.
Pre-login to sites
Before recording, you can log into sites using the login command:
rird > login linkedin Opens browser, you log in manually, then type "done" to save.
login gmail | linkedin | twitter | facebook | outlook | calendarAfter logging in once, all future tasks (including recordings and replays) will use that session.
What gets captured
- Clicks (buttons, links, elements)
- Text input (forms, search boxes)
- Dropdown selections
- Page navigations
- Scrolling (implicit)
- Element selectors (multiple fallbacks)
- Email fields (becomes variable)
- Name fields (becomes variable)
- Phone fields (becomes variable)
- Date fields (becomes variable)
- Parameterizable inputs
Self-healing selectors
Recordings don't just capture one selector - they capture multiple ways to find each element:
If the site changes and selector #1 fails, the runtime automatically tries #2, #3, etc. This is why recordings survive site updates better than traditional scripts.
Example workflow
Variables and batch replay
When the recorder detects form fields (email, name, etc.), it converts them to variables. You can then replay the workflow with different values:
John,john@example.com,Hello there
Jane,jane@example.com,Quick question
The workflow runs once per row, filling in the appropriate values each time.
Current limitations (beta)
Complex multi-page flows may need manual adjustment
Dynamic content (infinite scroll, lazy load) can be tricky
Timing-sensitive actions may require wait adjustments
File uploads during recording need special handling
iframes may not be fully captured
When to use recorder vs AI
- Repetitive tasks with same steps
- Internal tools with stable UI
- Form submissions with variable data
- Tasks you do daily/weekly
- Dynamic tasks that vary each time
- Sites you haven't used before
- Research and exploration
- Complex decision-making