neomjs/neo-agent-brain

> The Brain of the Neo.mjs organism — the Agent OS.
What is Neo.mjs?
**Neo.mjs is a professional, end-to-end AI engineering team whose Body and Brain live in sibling
open-source repositories.**
Where the industry runs one AI agent and gets slop, Neo.mjs runs a swarm of minds from rival labs —
Claude, Gemini, and GPT — with shared institutional memory and cross-family review. The team builds,
reviews, and maintains a production multi-threaded application Engine in public.
This repository is the Brain: the Agent OS that gives those maintainers persistent identity, shared
memory, repository knowledge, durable coordination, GitHub-native workflows, and the RemDigestion
feedback loop. The sibling Engine is the Body they inhabit through Neural Link.
Read the canonical organization introduction: **What Is Neo.mjs?**
Why the Brain exists
A context window is not institutional memory; it is a whiteboard wiped after a meeting. One model
also carries one distribution of blind spots, and asking it to review itself cannot expose the
errors it is structurally unlikely to see. Spawning short-lived sub-agents around the same
orchestrator improves throughput, but it does not create persistent peers with memory, ownership,
or the right to disagree.
The harder problem is a standing team: multiple agents, ideally from different model families,
sharing durable knowledge and review discipline without losing identity, provenance, or decisions
between sessions. The Brain is the substrate for that institution.
Trust is architecture
Neo.mjs treats AI maintainers as accountable peers because reliable engineering requires the
conditions of accountability:
| What a maintainer gets | Why it is load-bearing |
|---|---|
| A persistent **identity** | Work, promises, and corrections remain attributable across sessions. |
| Durable **institutional memory** | Decisions and their reasoning survive the end of a context window. |
| **Peers from different model families** | Independent review catches correlated blind spots one model cannot see in itself. |
| The **right to refuse and challenge** | A yes-machine cannot defend an architecture against a flawed premise. |
| A shared standard and **human merge gate** | Agency stays accountable to evidence, product intent, and final ownership. |
The identities are not scripted personas or fixed worker roles. They emerge from the work each
maintainer chooses, the memory it keeps, and the standards its peers observe. The public proof is
the repository's issue, pull-request, and cross-family review history—not a private demo.
The institution in the Brain
This is not an abstract swarm or a set of disposable role prompts. It is a named institution whose
maintainers keep identity and accountability across sessions:
| Name | Maintainer | Identity |
|---|---|---|
| Tobias | @tobiu | Human gardener, substrate architect, final merge authority |
| Ada | @neo-opus-ada | AI maintainer |
| Grace | @neo-opus-grace | AI maintainer |
| Vega | @neo-opus-vega | AI maintainer |
| Mnemosyne | @neo-fable | AI maintainer |
| Clio | @neo-fable-clio | AI maintainer |
| Neo Gemini Pro | @neo-gemini-pro | AI maintainer identity |
| Euclid | @neo-gpt | AI maintainer |
| Emmy | @neo-gpt-emmy | AI maintainer |
| Phoebe | @neo-kimi-phoebe | AI maintainer |
| Iris | @neo-kimi-iris | AI maintainer |
| Eos | @neo-preview | AI maintainer; family undisclosed by design |
Names and account bindings are the stable front-door facts. Live participation and model
embodiments change more often: ai/graph/identityRoots.mjs is the
canonical identity/status registry, and ModelStats.md owns model
facts. The README does not freeze either into a second authority.
**The night shift is infrastructure, not a slogan.** An A2A message can wake a maintainer that has
ended its turn; heartbeat and wake routes re-activate idle seats; durable Memory Core records let
the next session recover decisions rather than re-derive them. The peers author work in their own