Full-throttle agentic coding
The hands-on end. Run 40+ coding agents at once across tearable tabs, with an attention inbox that surfaces the one that needs you — live on desktop, web, and phone, with full history and ask-first permissions.
Multitudal is a suite of open-source tools for coding with AI — from full-throttle, hands-on agent driving to hands-off, walk-away autonomy, with the oversight to keep it all pointed the right way. However in-the-loop you want to be.
Two ways to ship with AI — drive it yourself or let it run — and one place to keep the whole thing on course.
Sits above both flows — links work to goals and keeps executive guidance over everything being built, whoever's building it.
Built for hands-on, high-intensity agentic coding — run 40+ agents at once across tearable tabs, with an attention inbox that flags the one that needs you, live on every device.
Fully hands-off. Drop a task in the hopper and walk away — Codey burns tokens until the work is polished, production-grade, and tested, then hands it back merged.
Each stands alone and ships as open source. Use one, or run all three across the whole spectrum.
Full-throttle agentic coding
The hands-on end. Run 40+ coding agents at once across tearable tabs, with an attention inbox that surfaces the one that needs you — live on desktop, web, and phone, with full history and ask-first permissions.
Executive guidance over both flows
The meta-orchestrator. Links work to strategic goals, hands tasks to humans or agents, learns from history to sharpen estimates, and keeps a clear view over everything being built — whether it flows through Agnes or CodeyBox.
Put work in. Get shipped code out.
The hands-off end. Drop a task in the hopper and walk away — Codey runs coding agents in isolated microVMs, burning tokens until the work is polished, tested, and production-grade, then audits and merges it onto your branch.
Most AI coding tools assume one developer, working alone. Multitudal is built for teams — so the whole group works alongside the agents, and each other.
Share a live session
Hand a running session to a teammate — or let them hop in and take over when you need to step away. Same agents, same history, no re-explaining.
Watch the queue together
The whole team can see the task queue, follow every run, and read the outputs as they land. No black box, no single gatekeeper.
One shared source of truth
Goals, health, and status in one place, with a view for every role — so developers, managers, and agents all stay on the same page.
Same work, three ways to run it — dial in exactly how much you want to be in the loop.
With Agnes, drive dozens of agents at once. Tearable tabs, an attention inbox, and every session live on whatever device you're on.
With CodeyBox, drop it in the hopper and go. Isolated microVMs, ask-first permissions, and an audit gate mean it's safe to leave running.
With JobTrack above both, work ties back to goals, health is visible at a glance, and the status report writes itself — no matter who's building.
No seats, no lock-in, no black boxes. Read the code, run it on your own hardware, and reshape any part to fit your team.
MIT licensed & free
Every tool ships under the MIT license, at no cost. Fork it, extend it, ship it — no strings, no per-seat bill.
Composed of plugins
All three are built from modular plugins. Edit, tweak, or replace core functionality yourself — swap in your own pieces and keep everything you find useful.
You stay in control
Isolated VMs, host-enforced egress rules, and audited merges. Give AI the work without giving up the keys.
Same tools either way. Self-host for free, or bring your own tokens and let us take care of the rest.
Free and open source, forever.
Managed by us. No ops, no upkeep.
Grab the source, run it yourself, and let the tools take the busywork. Then go grab lunch.
Explore the products