internOS
An open-source framework for human-agent collaboration.
AI agents lose context the moment a session ends, and humans and agents working together have no shared source of truth. Coordination falls apart across sessions, tools, and teammates.
internOS is an open-source framework for human-agent collaboration. It coordinates work through on-disk state files, so context survives across sessions and across agents. It runs this studio's own day-to-day operations.
How it works
- 01
Define the workstream
Each workstream gets on-disk state: a brief, status, decisions, and memory that any human or agent can read.
- 02
Work against the state
Humans and agents claim tasks and update status from the same files, so everyone sees the same picture.
- 03
Context survives
When a session ends or an agent hands off, the next one resumes from the files, with nothing lost.
What it does
- On-disk workstream state: brief, status, decisions, memory
- Task coordination between humans and AI agents
- Context that persists across sessions and across agents
- Portable: works with Claude, Codex, and other agents
- Open source, in real use running this studio
What stands out
Open source and in real use, not a demo.
Coordinates projects, tasks, and memory between humans and AI agents.
Context persists across sessions and tools.
Open source at github.com/poktalabs/intern-os.