How it works
Rational Trains is a multi-agent factory with customized CI/CD and cloud runs. You groom the work; it builds, checks and ships it, and tells you how it went.
Trains and cars
A release is a train. Each task is a car.
- Train
- A release. It leaves with its cars, stops at every station and arrives as a new version.
- Car
- One task: an issue you groomed. Each car is built in its own container, so cars never step on each other.
- Station
- A stage every car and every train passes: build, review, checks, release.
- Schedule
- No ship button to babysit. The train runs its stations on its own and reports to your phone.
- Manual tail
- Anything that truly needs your hands is listed when the train arrives, so nothing waits silently.
The steps of every car
You own the first two steps. The factory owns the rest.
StepsIdea → release
- 01IdeaYouAn issue, written from anywhere. A phone is fine.
- 02GroomingYouEpic, tasks, acceptance criteria.
- 03BuildFactoryAn agent works in its own container.
- 04Cross-reviewFactoryA different model reviews the change.
- 05TestsFactoryBuild and tests pass, or nothing ships.
- 06PreviewFactoryEvery pull request gets its own link.
- 07ReleaseFactoryBehind a feature flag, on for everyone.
Built like a good team
The factory follows the practices of a disciplined engineering team, and learns the same way.
- Parallel agents
- Several agents build cars at the same time. They use the factory’s own notification system to announce merges and changes, so nobody builds on stale code.
- Merge conflicts
- When parallel cars touch the same code, the factory resolves the conflict itself.
- Isolated runs
- Tests, lint and Prettier run in isolation on every change, on your machine or in the cloud, so a result never depends on whose machine it ran on.
- Critical bugs
- A critical bug found while a feature is being built is fixed in the same train, before release, not filed for later.
- Retrospectives
- Every train ends with a retrospective. What it finds goes into how the next train runs, so the factory learns from its own work.
Customized CI/CD and cloud runs
The pipeline fits your project, not the other way round.
- Pipeline
- Set up for your project, not a template: your checks, your rules for what may merge, your release steps.
- Agents
- Multi-agent: several AI agents build and review each car. No single model has the last word.
- Cloud runs
- On Google Cloud when you want the work off your machine, or locally when you do not.
- Delivery
- A preview link for every pull request, then a release behind a feature flag.