A software team of AI agents, on your own machine.
Give Mastwright a task. Its agents plan it, build it, review it and test it. You read every change and decide what ships.
From plain words to a change you can check
-
1
Write the task
Describe the work and what done looks like. For a bigger job, plan it first and let the planner split it into tasks.
-
2
The right agent builds it
Backend, frontend, mobile or DevOps. Each task works in its own copy of the project, so tasks can't overwrite each other.
-
3
Reviewed and tested
A separate reviewer reads the change and the test stage runs it. Failures go back to the agent with the reasons.
-
4
Merged and on Test
Work that passes is committed to your project's git history, and you can deploy it to a Test environment and try it.
-
5
You decide
Read the diff, then push it. If it isn't right, revert it and send the task back with your reason.
Your machine, your models, your decisions
It runs where you put it
Install Mastwright on your own Linux machine or with Docker Compose. Your code and your projects stay there. What leaves is only what goes to the language model you choose.
You choose the models
Use a cloud provider with your own key, or a local model on your own hardware. Different roles can use different models.
Agents don't deploy to production
Agents build, test and commit. Sending anything to production stays a person's decision, and people get their own logins and roles.
Upgrades that can roll back
You choose when to upgrade. Each upgrade backs up first, checks the result and rolls back by itself if something fails.
A mastwright built masts
The largest masts on a ship were too big to cut from one tree. A mastwright shaped several timbers to fit around a central core, then bound them with iron hoops into one strong spar. Our logo is that mast seen end-on: separate parts doing one job.
Tell us what you build
We're setting Mastwright up with a few companies at a time. Tell us what your team builds and where you'd want it to run.
hello@budjik.com