Mastra Factory - From issue to production, run by agents.
by•
Mastra Factory is an open source, agent-powered software delivery environment. It combines persistent coding agents, repository workspaces, issue intake, planning, implementation, and pull request review in a web application you control.
Replies
so many open issues? how can you call this finished?
@gagan_arora3 Mastra has 1.6m weekly downloads, Mastra Factory is only part of the open source repo.
Good product. I used the instructions here: factory.mastra.ai
@danvk thanks for trying it out! Let us know if you have any questions.
@danvk thank you! what should @mastra build/improve/fix from your perspective?
@danvk also if you find a bug or have an idea, open an issue here - contributions welcome!
builders would love this ngl
@shushantlakhyani Builds, designers, product manager, anyone. It makes sense to talk about Mastra Factory in the context of a software factory but, remember, it can be adapted for any kind of factory. For example, a content factory!
@shushantlakhyani @bookercodes huge opportunity
@shushantlakhyani exactly! help us spread the word on X, repost this
@shushantlakhyani what should @Mastra build/improve/fix from your perspective?
This is awesome... can't wait to try it!
@foobarfighter let us know if you have questions!
@foobarfighter great! you can learn more at mastra.ai/factory and view the source code on github.com. enjoy!
congrats!!
@hhendrix thanks!
@hhendrix thank you! help us spread the word on X, repost this
This seems like a cool product. All the best for the launch folks!
@sumitsaurabh927 thank you for your support! absolute fan of what you're building at @Lingo.dev btw cc @vrcprl @maxprilutskiy
oss ftw!
star the repo! github.com/mastra-ai/mastra
the hard part of issue to production is not the production half. it is that an issue is usually an under specified sentence, and the first thing any agent does is decide what it meant, silently.
a person picking up a vague ticket goes and asks. that ask is the highest value step in the whole loop and it is also the most awkward one to automate, because the agent has to be willing to stop and look unhelpful.
so the thing i would want surfaced in the run is not the diff. it is the interpretation. what did it decide the issue was asking for, written down before it touched anything. if that line exists i can review a hundred of these quickly. if it does not, i am reviewing a hundred diffs and i will stop after four.
@rabnoor_s Yeah!
#1 well earned 🥇 issue to PR with review built in is the right loop. how many agents run in parallel per repo?
@petrkovacik you can kick off as many agent sessions as you want. They run inside sandboxes.
We have run 50+ at one time across the org.
@petrkovacik thanks for your support! also curious how many agents do you run in parallel? fun fact: 80% of the community run up to 5 agents, according to this thread in /p/claude.
kicking off an agent overnight and just checking progress in the morning is such a nice feeling, like actual delegation instead of babysitting a chat window all day
@salvatoresng exactly, best feeling ever. help us spread the word on LinkedIn, repost this
@salvatoresng oh and btw @bookercodes and @suhdudeitsabhi are hosting a workshop this Thursday, September 17, at 4 PM UTC. they'll build a Factory, connect it to a real repo, and show how agents move an issue from initial investigation to a reviewed PR.
register here: https://luma.com/mastra-gogz