What is Farmslot?
Farmslot is a local operating system for supervised agentic engineering: a way to scale one engineer's judgment across many agent workers, projects, machines, and review loops.
It is the control plane around coding agents, combining current and emerging capabilities for roadmap memory, backlog refinement, dispatch, isolated slots, runner orchestration, live observability, recipe evidence, cross-runner review, and human approval.
Why the name? I know. It stuck: agentic dev farming across many isolated slots. Naming is harder than scheduling the agents.
One sentence
Farmslot turns many coding agents and project worktrees into an observable, evidence-backed engineering workflow that a human operator can still control.
The core problem
AI can generate code faster than teams can verify behavior. The next bottleneck is not typing; it is coordination, memory, validation, and trust.
A diff can look right while the app is still wrong. Farmslot changes the review question from:
Did the agent say it tested this?
To:
Can I inspect the exact run, evidence, reviewer findings, and approval gate?
The operating loop
The canonical loop runs from intent and memory, through backlog refinement, dispatch, isolated execution, observation, validation, review, human approval, and back into improvement data.
See Operating loop for the source-of-truth model. A team can adopt the loop incrementally: recipes and evidence can be useful before the full fleet harness; the full product scales the same trust loop across projects, machines, models, and operators.
A key design goal is low-friction project import. Today that means a small amount of pool/project configuration; over time, Farmslot should provide prompt-assisted import flows that inspect a repository, propose hooks, and make a new project dispatchable with minimal manual setup.