Your AI Startup, Complete with Org Charts and Kanban

What it does

OpenOPC is a framework that treats complex projects as the work of an entire AI company rather than a single agent. Given a goal, it automatically designs an org chart, staffs roles with specialized AI employees, and runs them through a kanban-style workflow with dependency tracking. A built-in learning system attributes outcomes to specific roles and compiles shared playbooks, so future “hires” start with institutional memory.

The interesting bit

The unusual angle is the organizational realism: it models reporting lines, role ownership, and a work-item state machine that handles parallel execution through a DAG while escalating blockers to humans only when they exceed the team’s authority. This is less “agent swarm” and more simulated enterprise, complete with delegation, review cycles, and HR-style onboarding.

Key highlights

• Operates in two modes: Task Mode (single-agent, LobeChat-like workspace) and Company Mode (full multi-agent organization)
• Self-Built mechanism drafts org charts and recruits AI employees from a talent pool, reusing veterans or hiring fresh roles as needed
• Self-Run orchestration uses a kanban state machine with dependency DAGs, enabling parallel work items and structured handoffs
• Self-Grown learning resolves feedback into per-role evaluations and promotes recurring lessons into shared playbooks
• Ships with both a React + Phaser Office UI and a full CLI for interactive and scripted usage

Verdict

Worth exploring if you need to automate complex, multi-domain projects that benefit from role specialization and process. Skip it if you just want a lightweight wrapper around a single LLM call.