How small fixes get done today
People keep small fixes in an issue tracker. GitHub Issues or Linear holds the bug report or the idea, and that text describes the work. It just sits there until you are free to do it. Then you open a coding agent and write a prompt for that item, one that repeats the issue, says what to change, and tells the agent how to decide the work is finished. The next item needs another prompt written the same way.
When the agent replies that it is done, the change is on a branch you still have to run yourself, so you check out that branch, start the dev server on your machine, and then click through the page there to see whether the fix holds.
The same backlog in Rondar
Every bug or idea goes in as a todo. The todo is the request in plain words, so it can stay in the list until you choose to run it.
Then you select several todos and run them, and Rondar gives each todo its own session, started from the todo's text and run in an isolated copy of the repository. How many sessions run at once depends on the plan; any todo past that waits, and Rondar starts it as soon as a slot opens up.
When the session has made the change, Rondar runs the app with that change. You click through it in the browser, beside the conversation. If a page is wrong, you continue the conversation and the agent keeps working in that same session. Rondar opens a pull request for the session. It monitors the checks and repairs failures.
After you approve the changes for the release, Rondar merges them together and runs the tests. The release can deploy to staging and to production.
What changes for the developer
The todos you saved are the queue of work. You choose which items run, and each one gets its own session. A session that is still waiting starts when a slot is free. Rondar sends a browser notification or an email when a session finishes or asks a question. You can set those alerts per project.
You click through the app with the change before you approve the session for release. Rondar merges the approved changes together and runs the tests on that combined result.
How to try it on your own project
Fill in the waitlist form on the Rondar website to request an invitation to the invite only alpha, which includes a free pilot. The pilot covers 14 days, and after that plans start at $20 a month.
Connect the GitHub repository for the project. We clone it and prepare an isolated workspace where the coding agent works. Connect the coding agent you want to use, such as OpenAI Codex or Claude Code. The session runs on your provider account and stays within that account's limits. You can connect more than one account.
For the app to run in the session, make it listen on the PORT variable and bind to 0.0.0.0. Add example environment values and seed data, so those screens have data when you click through them.
Add the bugs and ideas you already track as todos. Select several and run them. We make one session per todo and start the next when a slot is free.
When a session finishes, click through the app with that change and approve the ones you want. We merge those changes together and run the tests.