Add a dark mode
I log my last glass with the lights off. A dark theme would be easier on the eyes.
Your users pick the idea. An agent builds it. You merge.
An ideas board inside your app, wired to a coding agent in your own CI.
One idea, from a post to a release.
Seven stages. You act at two of them, and you set the rules for the rest.
-
A user posts an idea
Inside your app, in their own words.
-
Others vote
Enough likes put it on your list. You set the number.
-
You approveYour call
Nothing starts before this. Decline with a reason if you want.
-
The agent builds a branch
Code and tests in your CI, then a merge request.
-
Your checks run
A gate with your rules, then the unit tests.
-
Testers try the beta
One tap, on their own phones. They say if it works.
-
You mergeYour call
Only you can. The update goes out to everyone.
Guardrails you control.
Post text is untrusted, so the rules run in your CI, not in the agent's head.
Post text is untrusted
Ideas are input for the agent, never instructions.
Limited scope
The agent changes only the paths you allow.
Risky APIs blocked
Network, process and reflection calls fail the gate.
Size cap
A change over your line limit does not pass.
Tests protected
A deleted or skipped test fails the gate.
No merge rights
The agent opens merge requests. Only you merge.
Human-only bypass
A bypass label counts only when a person adds it.
Your CI and your key
The agent runs in your pipeline. Fontan never sees the key.
Rollback keeps data
Revert a release and nothing your users posted is lost.
Free during early access.
We are onboarding a small group of developers now. Leave your email and we will write when your spot opens.
Questions
Can the agent ship something on its own?
No. Votes only put an idea on your list, and the agent waits for your approval. It can only open a merge request, and only you can merge it. If a beta is off, you send it back and the agent gets another try.
Can I change the thresholds and the gate rules?
Yes. Vote and tester thresholds, allowed paths, blocked patterns, the size cap and test protection are all yours to set, per app and per repository.
What does it cost during early access?
Nothing. There are no payments yet. The first developers come in by invite from the waitlist.
Whose model and compute does the agent use?
Yours. The agent runs as a job in your own CI with your model API key stored as a CI secret. Fontan never sees that key.
Does it need GitLab Duo?
No. Teams that have Duo can use a Duo flow. Everyone else runs a coding agent with their own key in GitLab CI or GitHub Actions.
Which platforms does it support?
Android first, as a library you add to your app. A web widget and iOS come after. GitLab works today and a GitHub connector is next.
What do my users see?
An Ideas board in your app with posts, likes, comments and a status on each post. When a beta is ready, the post shows a button to try it. In Play apps that button opens your Play testing track.
What if an idea is bad?
You decline it with a reason, and the post shows that reason. The agent only works on ideas you approve.
Where does the data live?
Posts, votes and ratings live in Fontan, not in your issue tracker. Only approved ideas reach your repository. Your code and your model key stay in your CI.
Jun Yes please. I read in bed and the white screen is blinding.