Run many agents across many project types, computers, and model subscriptions. Watch and steer them while they work, then trust the result through visual validation, executable recipes, and proof artifacts instead of final claims.
curl -fsSL https://raw.githubusercontent.com/deeeed/farmslot/main/install.sh | bash
farmslot doctor.The hosted Command Center is published as a static page at farmslot.io/cc. For today’s plain local gateway, run farmslot up and use the local Command Center URL it prints; browsers can block HTTPS pages from opening a plain ws://localhost gateway. The hosted page remains available for secure gateways and the upcoming secure local bridge.
farmslot up
Starts your local gateway, opens the local Command Center fallback, and gives you the hosted link to try once the gateway is reachable from the browser. Use Doctor to verify gateway, projects, slots, browsers, simulators, ADB, and capture-helper.
curl -fsSL https://raw.githubusercontent.com/deeeed/farmslot/main/install.sh | bash
Installs prerequisites, configures Farmslot, then guides you to launch the gateway and connect Command Center.
farmslot.io/cc is deployed with the docs site, but GitHub Pages redirects to HTTPS. If your browser blocks the plain local gateway, use the fallback URL printed by farmslot up, usually http://localhost:7777.
Farmslot is the product layer around agents: a framework for many project types, a control plane for coordinating the fleet, a cockpit for watching and steering work, and a proof-first recipe layer for visually and mechanically validating the result.
Coordinate projects, machines, slots, runners, hooks, run state, artifacts, and memory through one framework.
Open docs →Use Command Center as the cockpit: live terminals, decisions, app state, artifacts, and intervention while the run is still moving.
Open docs →Validate visually and mechanically with recipes at the core: real actions, assertions, screenshots, traces, HUD intent, and proof artifacts.
Open docs →Bring mobile apps, web apps, extensions, CLIs, services, desktop apps, and headless systems into the same operating loop.
Open docs →Turn multiple model runners, computers, and subscriptions into coordinated parallel engineering capacity.
Open docs →Keep approvals, workers, runs, and terminals visible when you step away from the workstation.
Open docs →Turn failures, flakes, and review comments into better prompts, docs, recipes, checks, and adapters for the next run.
Open docs →These examples show what Farmslot lets you do: coordinate active agent work, turn validation into reusable proof, and supervise different project types through one operating loop. The same workflow scales from live terminals to review gates, mobile supervision, and repeatable validation.
Command Center gives you a live cockpit for parallel agent work: see each slot, watch terminals move, steer a run before it drifts, and keep enough context to make a decision while the work is still active.
Gateway intelligence turns fleet state into an operator assistant: ask what is running, which slots need attention, or where proof artifacts live without manually scanning every terminal and run card.
Real Command Center screenshotReady gates make agent output reviewable before it becomes your code: inspect the package, app behavior, validation results, diff, and quality checks, then approve, request another review, or send feedback to keep steering.
Validation trailRecipes make validation repeatable: instead of trusting a final claim, you can replay the actions, assertions, screenshots, videos, traces, summaries, and manifests behind the result.
Real app screenshotThe Companion app keeps the operating loop available away from the workstation: check active runs, inspect proof artifacts, see slot and terminal state, and stay close to approvals when you step away.
Concept diagramFarmslot is not tied to one stack. Model each project type with hooks, resources, and recipes, then supervise them through the same dispatch, validation, review, and retrospective loop.
Farmslot treats validation and retrospectives as part of the product loop. Runs create evidence, review finds gaps, retrospectives capture durable lessons, and those lessons improve recipes, prompts, docs, and adapters.
Read the feedback loopImport a project with hooks, add a runner manifest, adopt recipes where behavior needs proof, then use retrospectives to keep improving the system.