SayeOS Core · Free · Coming soon

A governed workspace for intelligent work.

Connect the models you choose, work across people and software workers, use Python for real execution, and preserve the context, state, evidence, and history that let work continue over time.

Core is not yet publicly distributed. Join the preview for release updates and early access. Signup opens your email app to hello@sayeos.org.

What SayeOS is

Persistent work that does not depend on one model, system, or connection.

SayeOS is a governed workspace where people, models, Python workers, APIs, and software services contribute to persistent work, without making one model, one system, or one connection method the center of the architecture.

Persistent workstreams

The work continues across sessions, contributors, and model changes.

Agnostic by design

Systems, models, and connection methods can change without redefining the work.

Python execution

Models reason. Python executes defined work.

Public Constitution

The governing foundation can be read.

The persistent workstream

Contributors change. The work persists.

WORKSTREAMIntentContextStateEvidenceDecisionsHistory Person Model A Model B Python worker API Database Software service Reviewer
Models and tools can change without forcing the work to start over.

A core principle

System, model, and connection agnostic.

SayeOS does not make one application, one provider, or one connection method permanent. The work stays independent of the tools contributing to it.

System agnostic

SayeOS works around existing systems rather than requiring one application to become the source of everything. Accounting systems, CRM, GIS, document systems, databases, local files, and operational software keep their authority. SayeOS coordinates work around them.

Model agnostic

The work is not owned by one model provider. Different models contribute according to task, cost, capability, privacy, environment, and policy. Changing providers does not force the workstream to restart.

Connection agnostic

SayeOS does not assume one universal way to reach systems, models, tools, or workers. Connections include APIs, local tools, files, databases, approved services, Edge mediated access, and defined interfaces. No single connection method is permanent or universal.

Why it matters

Models change. Systems change. Connections change. The work should survive.

A better model arrives. A provider is replaced. A system changes. A local capability replaces a remote one. An API is swapped for another approved interface. Through all of it, the workstream keeps its context and history.

Persistent workstream CONTEXT · STATE · HISTORY BEFORE AFTER Model A System A API A Model B System B Local connection
Models change. Systems change. Connections change. The work should survive.

The work should outlast the tools contributing to it

What you can do with Core

Real work, held together over time.

Individual

Research a major decision

Maintain sources, assumptions, alternatives, calculations, open questions, and decisions across months of work, from a relocation to a career change.

Small team

Prepare for fundraising

Founders, finance, model analysis, Python forecasts, source data, assumptions, and reviewer input in one workstream.

Builder

Create a reusable Skill

Turn Python code, inputs, permissions, validation, and review boundaries into a capability others can inspect and reuse.

Continuum

Understanding that grows without losing where it came from.

Continuum preserves understanding over time without stripping away where it came from, what it meant in context, how well it is supported, or who has permission to share it.

Original voice and normalized meaning are both kept. Confidence changes as evidence arrives. Disagreement is preserved rather than flattened. Nothing becomes public by accident.

  • Original voice and normalized meaning
  • Source, contributor, and context
  • Confidence that changes with evidence
  • Disagreement kept visible
  • Explicit sharing boundaries

Models reason. Python executes.

Python is a first class execution layer. Python workers execute defined analysis inside the workstream, across data, finance, GIS, documents, APIs, and databases. Execution stays bounded by security, dependencies, and permissions.

Extend through Context and Skills.

A Context Pack carries governed knowledge with known provenance. A Skill is a bounded, executable capability with defined inputs, allowed tools, permitted actions, and review. Use Saye and Marketplace content, build your own, and publish eligible offerings.

Public Constitution

Consequential work runs under rules you can read.

SayeOS is governed by a public Constitution. Human authority stays primary. Intelligence may propose but not decide. Authorization is explicit and revocable.

Why we built it

We built the system we wanted to use.

Capable models were not enough. The work needed to persist. Existing systems needed to keep their authority. Models needed to be replaceable. Connections needed to stay flexible. And the rules governing consequential work needed to be visible.

Coming soon

Join the SayeOS Core preview.

Building custom governed AI systems for an organization? See SayeOS Enterprise.