AIOS

An AI-powered operating system for agentic teams.

Structured agent workspaces + a shared collective brain. MIT-licensed, self-hosted. You decide what leaves your machine.

Open source MIT licensed 2026
01 · The problem

Your organization already has a brain.

It's scattered across Slack threads, stale docs, meeting notes, and the person who left six months ago. In agentic teams, context bloats per person, tools entangle, and speed in the wrong direction is waste.

AIOS is the missing layer that makes collective intelligence queryable, actionable, and alive.

02 · Why now

The coordination tax is crushing organizations.

TeamsCompaniesOrganizations
03 · Principles

Four design commitments.

P1

Modular

Each organ system is independently deployable. Adopt what you need.

P2

Open source

MIT-licensed and self-hosted. Clone per person, run the brain on your Postgres.

P3

Permissioned

Every action is policy-governed. Teams define autonomy vs. human approval.

P4

Privacy-first architecture

Individual data stays personal. The brain receives only what members push.

04 · Architecture

Individual agents → collective brain → surfaces → actions.

05 · The system

A workspace for individuals, a brain for the team.

Your workspace

Numbered folder spine, governance, validators, and harnesses — one per person.

Shared memory

Tier-tagged pushes to one brain. Query in plain English across team context.

Nothing leaks

Private → team → external. Admin-tier rejected at the API boundary.

06 · Runtime-agnostic

One interface for all your agentic work.

Bring your own agent. Extend with MCP. Drive everything from a GUI — no terminal required.

ClaudeClaude CodeCursor CodexOpenCode+ any MCP client
07 · How it flows

A continuous loop, not a pipeline.

Work surfaces from every workspace, enriches the shared brain, and insights return to each individual — tightening with every cycle. Share a skill to the team repo and everyone adopts it.

08 · Vision
"What Claude Code did for individual agent productivity, AIOS does for team agentic productivity."

Every context push makes the system smarter; every query pulls the right signal back. Agents take on execution; humans run orchestration.

09 · Anatomy

Eight organ systems.

01
Knowledge
✅ Shipped
02
Ingestion
✅ Shipped
03
Context
🟡 Partial
04
Actions
🟡 MVP
05
Identity
✅ Shipped
06
Policy
🟡 Partial
07
Audit log
✅ Shipped
08
Feedback
🟡 Partial

Read the full reference →

10 · Honesty

The hard problems.

01

Context overload

Signal over noise — a brain that pages you about everything is worse than no brain.

02

Contribution incentives

Social design matters as much as technical design.

03

Governance

Building coordination requires coordinating the builders.

11 · Roadmap

Where this is going.

Shipped

  • Runtime-agnostic integrations
  • Tiered access governance
  • Agentic maturity instrumentation
  • Open-source design system

In progress

  • Unified LLM cost rollup
  • Two-way PM sync
  • Slack bidirectional integration

Planned

  • Agentic learning journeys
  • Public live roadmap board
  • Finance & runway snapshots
12 · Get started

Up in three commands.

bash
$ npx aios-init my-workspace
$ aios run meeting-synthesis --input ./transcripts/
$ aios push --tier team
AIOS

Built for what comes next.

Every institution today was designed for human actors. We are building the infrastructure for human-agent ensembles.

MIT License · Built by John Ellison & Chetan Nandakumar

1 / 14