Batur Yilmaz Arslan develops software for agent coordination and language learning. His Agent Office project gives multiple coding agents a shared workspace with explicit scheduling, messages, task state and operational controls.[1][2][3]
Start with Agent Office (Slack for AI Agents) – Similar to Grok Bot but older →
Contributions
Give agents a managed workspace
Agent Office builds on the Pi agent runtime with tick-based scheduling, priority queues and inbox communication. An office is configured as a workspace containing agents and shared operating rules. Its monitoring and watchdog features address the lifecycle around a run, including agents that stop responding or need new work.[2]
Turn collected words into learning material
Wordpecker combines user-curated vocabulary with language-learning lessons. Users can collect words from material they read or watch and revisit them through practice. The project connects generative features with a persistent personal vocabulary rather than requiring everyone to follow the same fixed word list.[1][3]
Ideas
The project documents a hierarchy of tool policies, heartbeat permissions and office-level configuration. Docker isolation is optional and differs from in-process execution. These settings determine what an agent can access and do; a collection of agent roles is not, by itself, an isolation boundary.[2]
Tools, writing & talks
Writing, research & talks
Start here
A Pi-based workspace manager for scheduling agents, sharing messages and controlling tools and execution.
Requires pnpm and configured provider credentials; Docker is optional. Read the architecture, office configuration and sandbox modes before the example teams. Check the permissions and tool policies for the tasks you want agents to perform.
Developer and maintainer[1][2]Original source
Apps
- Wordpecker
github.comA language-learning application built around a user’s own vocabulary collections.