Project Utopia
GitHub

Open source · MIT · Building in public

Speak once. Let the system orchestrate the rest.

Project Utopia is an open-source operating system for personal AI. Say what you need — it understands the goal, coordinates specialist agents in parallel, verifies their work, and returns concise proof.

“Plan the launch — research the market, draft the site, schedule the rollout.”

View on GitHubExplore the system

Early development — the architecture is real, the code is landing in public.

01 / Vision

One sentence in. A coordinated system out.

Most assistants answer. Utopia acts. A single spoken request fans out into parallel, verifiable work — and converges back into one concise result.

  1. 01

    You speak

    Say it the way you would to a colleague. The realtime voice layer captures intent — the goal, not keywords.

  2. 02

    Utopia orchestrates

    The durable core breaks the goal into work, picks the right specialists, and keeps the session alive until the job is done.

  3. 03

    Agents execute in parallel

    Implementation and review agents run concurrently — researching, writing code, recalling context, scheduling what comes next.

    • research
    • code
    • review
    • memory
    • schedule
  4. 04

    You get proof

    Results are verified before they reach you. The answer comes back concise, with evidence — an outcome, not a transcript.

02 / System

Every layer replaceable. The core durable.

Utopia is an operating system, not a chatbot: a small set of honest layers with clean seams. Swap the voice provider, swap the worker models — Hermes keeps the state, the schedule, and the standards.

  1. L0

    Voice interface

    Replaceable

    A thin realtime layer turns conversation into intent. OpenAI Realtime is one supported provider — the voice layer is a socket, not a dependency.

  2. L1

    Hermes orchestration core

    Durable

    The heart of the system. Hermes holds durable sessions, decomposes goals into work, routes it to the right specialists, and survives restarts mid-task.

  3. L2

    Specialist workers

    Parallel

    Implementation and review agents — Codex and Claude among them — execute concurrently and cross-check each other's output before anything ships.

  4. L3

    Utopia Memory

    Private

    Personal context lives in a store you control. Agents recall what matters without your data leaving your hands.

  5. L4

    Utopia Pulse

    Proactive

    Watches the schedules and signals you care about, and raises work before you think to ask.

  6. L5

    Approvals & verification

    Human-gated

    Consequential actions wait for explicit approval. Every result is checked, and what comes back to you carries its evidence.

03 / Open Source

Built in the open, from day one.

Project Utopia is early — and honest about it. No launch theater, no invented numbers. The architecture, the prototypes, and every decision land in a public repository under the MIT license.

Follow the repositoryMIT License

Current state

  • Realtime voice loopprototyping
  • Hermes orchestration corein design
  • Specialist worker poolprototyping
  • Utopia Memoryplanned
  • Utopia Pulseplanned

Status reflects the public repository, not a roadmap promise.

Join early

Help build the operating system for personal AI.

Star the repository to follow along, open an issue to shape the architecture, or send the first pull request. Everything happens in the open.

github.com/Larry-MG/Project-Utopia