ACME

Build, share, and run AI agents with your team.

Discover, clone, and run versioned AI agents — with shared memory, scheduling, sandboxing, and messaging channels.

01

Share agents with clones and shared memory

Clone an agent into your workspace and choose how memory is shared. Start from a proven agent without rebuilding context from scratch — then keep learning together.

02

A harness built for real work

Every agent runs on a powerful harness: Pi for conversations, scheduled jobs, just-bash workspaces, and an encrypted vault for secrets. You get production-ready runtime primitives without wiring them yourself.

03

Git and file first

Prompts, tools, schedules, and config live as files in a versioned bundle. Diff, pin, and roll back like code — so agent changes stay reviewable and reproducible.

04

Organizations, teams, and permissions

Share agents across your organization and teams with a clear permissions model. Give the right people access without spreading secrets or losing ownership.

05

Efficient by default

Agents run in the cloud on schedules instead of sitting idle 24/7. Pay for useful work — cron-driven turns, enable/disable control, and no always-on deployment tax.

06

Channel your agents

Talk to your agents where your team already is — Discord, Telegram, and other popular messaging apps. Bring the agent to the channel, not the other way around.