tessryn

Open source / Original kernel & OS

An operating system
for distributed intelligence.

Tessryn is being built for AI agents and distributed inference. Its kernel, native drivers and cluster services are designed together so qualified machines can share the work.

Build with us on GitHub Read the design

Apache-2.0 Public design & implementation backlog

Planned architecture: agent and model execution above native drivers and cluster services, supported by an original kernel with owned resources.
FIG. 01 Architecture in development

Design & implementation planning

No installable release, qualified hardware profile or measured performance results yet.

Roadmap

01 / Build with us

Good systems start
with contributors.

Kernel engineering, native drivers, distributed systems, inference and measurement. Choose a focused piece of the foundation.

  1. 01

    Understand the design.

    Start with the architecture, decisions and evidence gates.

    Explore the design
  2. 02

    Choose an open issue.

    Check dependencies and agree on the acceptance criteria before starting.

    Browse open work
  3. 03

    Make a focused contribution.

    Open a pull request with reproducible evidence and clear limits.

    Contribution guide

Get the source.

Development checkout · Python 3.11 or newer

git clone https://github.com/AG187-Lab/tessryn.git
cd tessryn
python3 --version
python3 tools/validate_design.py

These commands check the source documentation and implementation plan. They do not install or boot Tessryn.

Foundational work

Platform qualificationReproducible builds & evaluation

02 / Install & use

Your machine.
A clear starting point.

Development is open today. Running the operating system comes after the boot, hardware and release gates.

On your server

Target hardware / Qualification first

The first targets are exact x86-64 UEFI profiles with native networking, NVMe storage and one selected NVIDIA family. Installable images will follow the roadmap’s release gates.

There is no server installer to run today.

Server targets

On your Mac

Development host / Source checks today

Use your Mac to explore the design, edit the source and run the checks above. Native Apple M5 boot, GPU and transport support have separate qualification gates.

A native macOS installer or worker app is not available.

Mac roadmap

As a hosted cluster

Planned product / No public launch yet

The intended hosted service will provide a console for creating clusters, enrolling qualified machines and placing models. The operating system remains an independent open-source project.

Provisioning, enrollment and native APIs still need implementation and qualification.

Planned workflow

03 / The hosted experience

From machines
to a managed cluster.

The plan is a managed experience in the spirit of EKS or DigitalOcean: create a cluster, connect qualified machines and deploy a model.

A hosted management service will handle the setup. Tessryn will run on the workers.

Proposed workflow · Not available yet
  1. 01

    Create a cluster.

    Select a qualified profile and an explicit resource budget.

  2. 02

    Connect machines.

    Boot Tessryn and authenticate enrollment of eligible workers.

  3. 03

    Place a model.

    Verify artifacts, reserve memory and distribute resident shards.

  4. 04

    Use the cluster.

    Submit inference requests and bounded agent tasks through native APIs.

The public OS’s first product proof is a four-machine cluster, including a qualified wired laptop, executing one model larger than any individual worker can hold.

Follow the roadmap

The work is open.

Help build the foundation.

Contribute on GitHub