After too many rounds of "the agent says it's done and it's just a pile of ", we built the opposite: no_human proves the code it wrote is correct. You drop a ticket on the board (or point it at Jira, Linear or Monday) and it plans, writes the code, and opens a pull request - with the evidence that everything works!
Before that PR reaches you, the work is checked by a second model that never saw the coder's session and is told to assume the job is not done. They communicate until the reviewer is convinced everything is good to go (the num of attempts is limited to prevent an infinite loop).
If the coding agent deleted or weakened a test, a tamper guard stops the attempt. For bug fixes, the tests offered as proof have to fail on the old code and pass on the new.
Under the hood it's a team of agents: a planner, a coder, an independent reviewer in a fresh session, and a supervisor.
It's free and open-source, on your machine.
no_human has been building itself for about two months now. We are feeding ticket into our jira board, it pulls them from there and just work on them until eveything is done. We trust it enough that we've enabled auto approve mode so all PRs no_human ships - are approved and merged automatically (this feature hasn't been released yet).
Entire sprint's worth of work is done in 3-4 days, while we focus on the bigger picture - our strategy and the next steps.
It currently supports all Antropic and OpenAI models and also has a quite unintuitive process of working with local models as well (we are working on improving that).
It's open to contributions (we would love to see you use it and contribute!!!)
After too many rounds of "the agent says it's done and it's just a pile of ", we built the opposite: no_human proves the code it wrote is correct. You drop a ticket on the board (or point it at Jira, Linear or Monday) and it plans, writes the code, and opens a pull request - with the evidence that everything works!
Before that PR reaches you, the work is checked by a second model that never saw the coder's session and is told to assume the job is not done. They communicate until the reviewer is convinced everything is good to go (the num of attempts is limited to prevent an infinite loop).
If the coding agent deleted or weakened a test, a tamper guard stops the attempt. For bug fixes, the tests offered as proof have to fail on the old code and pass on the new.
Under the hood it's a team of agents: a planner, a coder, an independent reviewer in a fresh session, and a supervisor.
It's free and open-source, on your machine.
no_human has been building itself for about two months now. We are feeding ticket into our jira board, it pulls them from there and just work on them until eveything is done. We trust it enough that we've enabled auto approve mode so all PRs no_human ships - are approved and merged automatically (this feature hasn't been released yet). Entire sprint's worth of work is done in 3-4 days, while we focus on the bigger picture - our strategy and the next steps.
It currently supports all Antropic and OpenAI models and also has a quite unintuitive process of working with local models as well (we are working on improving that).
It's open to contributions (we would love to see you use it and contribute!!!)