KeelenAnswer — unattended work
I have no time for my side project. Can AI work on it for me?
Updated 20 August 2026
Yes, with one condition that most people skip: the project needs gates. Everyone who has tried leaving a coding agent running unsupervised reports the same failure, which is plausible code that nobody verified. The fix is not a better prompt, it is infrastructure: a test suite, a CI pipeline, and a rule that nothing merges while a check is red. With those in place, an autonomous loop such as Keelen can genuinely work a repository while you are at your day job, opening one tested pull request at a time for you to review when you get to it.
The realistic division of labour is that you keep the direction and the review, and hand over the hours. That is what was actually missing, and it is the part that does not need you.
What to do, in order
Accept the real constraint
The blocker on a stalled side project is almost never ideas or skill. It is a supply of uninterrupted hours you do not have. Solve for hours, not for motivation.
Build the gates before the autonomy
A smoke test, tests on the critical paths, and a CI workflow that runs on every pull request. Without these, unattended work is unverifiable and you will end up reviewing everything by hand anyway, which costs the hours you did not have.
Write your intent down where a machine can read it
A standing backlog of specific requests beats a chat session, because it survives you not being there. Vague items produce vague work whoever implements them.
Keep the work in small pull requests
One task per pull request keeps each change reviewable in the ten minutes you actually have, and reversible if it is wrong.
Review on your own schedule, not the machine's
Set the project to open pull requests and wait, or to merge only on green CI after a review window. Either way the queue waits for you rather than the other way round.
Why unsupervised agents get a bad reputation
The common experiment is to point a coding agent at a repository and leave it running. It produces a large volume of plausible code, some of it wrong, and no mechanism decides which is which. The lesson people draw is that autonomy does not work. The accurate lesson is that autonomy without verification does not work, which is also true of humans, and is why professional teams have code review and CI in the first place.
What the gates are
Five gates stand between the loop and your main branch. New tests are applied without the implementation first and must fail, so tests born green are rejected. An independent reviewer that shares no context with the run that wrote the code reviews the diff. Your own test suite runs on a clean checkout. Nothing merges over a red required CI check. And a review window stands before gated auto-merge, which you can replace with plan-review sign-off, manual merge, or branch-only mode.
What it costs to run overnight
The Indie plan is $29 a month: 3 projects and one on the 24/7 schedule. Keelen runs on credentials you connect: Anthropic Claude (a Claude Code subscription login or an API key, no Max plan required), OpenAI Codex, Zhipu GLM, or Moonshot Kimi. Token usage bills to your own provider account at cost. Keelen charges a flat monthly tier and never marks tokens up. If you already pay for a Claude, Codex, GLM, or Kimi plan that idles most of the day, this is mostly a way to use capacity you have already bought.
When it gets stuck at 3am
Failed runs are classified into 30 kinds. Infrastructure blips and provider limits retry on a budget and never count against your work. A real dead end becomes one plain-English card in your Needs-you queue with a recommended action, and the project pauses rather than burning quota.
When Keelen is not the answer
- Your project has no tests and you do not want any. Then unattended work cannot be verified, and you should not run it.
- You enjoy the coding and want to keep doing it. This buys hours you may not want back.
- The project is exploratory and changes direction weekly. A standing roadmap suits work with a known direction.
Hand the work to a loop
Connect a repository, write what you want in plain language, and review the tested pull requests that come back.
FAQ
Will it break my project while I am asleep?
It cannot merge over a red required CI check, and your own test suite runs on a clean checkout before that. If you want a harder guarantee, set the project to manual merge or branch-only, and nothing reaches your main branch without you.
How much of my own time does it still need?
Realistically: writing requests when you have new intent, and reviewing pull requests. Both fit in short sessions, which is the point. What it removes is the multi-hour block you never get.
Can it work on a project that has no tests yet?
Yes, and the first requests should be the tests and the CI workflow. Until those exist, keep the project on manual merge and review each pull request yourself.
Do I need a Claude Max plan?
No. Keelen uses whatever credentials you connect: a Claude Code subscription login or API key, OpenAI Codex, Zhipu GLM, or Moonshot Kimi.