SageOx

The hivemind for agentic engineering

Announcements

Introducing SageOx

The hivemind for agentic engineering

The SageOx Team

By The SageOx Team

In the agentic era, the engineering bottleneck has shifted. Execution is abundant. Alignment is scarce.

Agents produce code quickly — but they diverge just as quickly. Sessions are isolated. Prior decisions are forgotten. Patterns conflict. Speed compounds into drift.

You fix a bug with an agent. A week later, another engineer reintroduces it because the original reasoning wasn't preserved. You debate an architectural direction in a meeting. The next agent session scaffolds something that contradicts the decision — because it had no idea the discussion happened.

For a solo developer, this is friction. For a team, it becomes architectural entropy.

The Missing Layer

Traditional product teams preserve shared intent through documents, reviews, and process. Those artifacts accumulate over time and keep everyone pointed in the same direction.

But agentic teams operate through ephemeral sessions at machine speed. Context must be manually assembled. Alignment depends on individual discipline. Nothing persists unless someone explicitly writes it down. Retrofitting traditional processes doesn't bridge the gap — it undermines the speed that makes agents valuable in the first place.

What's missing is a layer that enables alignment at machine speed. We see it as a new category: agentic context infrastructure.

SageOx v0.1

SageOx provides agentic context infrastructure.

SageOx gives AI-native teams a shared memory that every human and every agent draws from automatically before they act. The loop is simple: context is no longer something individuals carry in their heads. It becomes infrastructure.

How it works

A continuous loop that turns ephemeral discussions and sessions into durable, searchable team context.

Capture

DiscussionsSessions

Structure

SummariesTranscriptsIntent

Consult

Agent sessionsPRsReviews

Agent sessions don't start from zero. They start from shared memory.

SageOx v0.1 is for teams that build products primarily through prompts.

It makes human–agent collaboration multiplayer by default through four core components:

1

Team Context

Agents are grounded in the artifacts that define how the team operates — principles, norms, architectural decisions, onboarding documents, constraints, and conventions.

2

Ledger of Work

SageOx records architectural and product intent as it emerges — in human discussions and in human–agent sessions — and structures it into durable, searchable history.

When two engineers debate in a meeting whether to use REST or GraphQL, that decision and its rationale become a queryable context for every future agent session. Similarly, when a developer collaborates with an agent to implement a feature, the reasoning behind the implementation becomes part of team memory, without anyone having to manually document it.

Disparate human discussions and isolated sessions become shared memory.

3

Ox CLI

Ox automatically primes every new agent session with relevant team context — recent decisions, architectural choices, active constraints. No manual assembly. Agents consult the hivemind before they act.

4

Web App

The web interface allows teams to organize members, connect repositories, and access the ledger through a UI.

Plowing in the Open: Ox CLI

Agentic engineering demands a fundamental evolution of how teams operate — roles, processes, and decision-making across humans and AI. SageOx's mission is to enable that evolution, with the operating model as the primary focus and the infrastructure in a supporting role.

We're leaning in: not open source, but open work. Signed-in SageOx users get a live, unfiltered view of how we build the Ox CLI — the debates, the wrong turns, the trade-offs, the moments where we change our minds. Not just what we ship, but how we think and co-create across humans and AI. The raw stuff. The ongoing stuff. The stuff that usually stays behind closed doors or private to our laptops.

We're convinced it's how the best agentic teams will operate: visible reasoning, inspectable decisions, compounding context. As we refine our own workflows and learn alongside other pioneering teams, we publish the patterns that emerge — in our operations, in the product, and in writing. If you're curious how agentic teams are actually figuring this out, this is a front-row seat.

Demo: From Ephemeral Discussions to Live Features

See it in action

A real workflow using the public Ox CLI project — every artifact is visible to logged-in users.

Live Demo Walkthrough

From Ephemeral Discussions to Live Features

SageOx in practice: using the public Ox CLI project.

1
Capture

Technical Meeting

A technical meeting is recorded and processed inside SageOx as durable team context.

2
Build

Context-aware Agent Implementation

With a single prompt, developer asks agent to implement the discussed feature. Their session is recorded with ox-session — reasoning and intent captured automatically.

3
Consult

Shared Context for the Team

The same session lives inside SageOx for every coworker — no context lost, no re-explaining what was done or why.

4
Ship

Code with Context

The code commit traces back to the decision that motivated it — origin, reasoning, and intent all linked.

Imagine operating at this speed for all of your technical and product decisions.

No manual context assembly. No re-explaining past reasoning. No alignment drift.

Get Started

Download the CLI:

Claude Prompt: Take a look at gh sageox/ox and install the cli

Create an account:

Claude Prompt: ox login

If your team builds primarily through prompts, we'd love to hear from you: feedback@sageox.ai

With Gratitude

First and foremost to Steve Yegge, who wrote the book on Vibe Coding and introduced the world to Beads and GasTown. There is a straight line from several of our conversations to today's launch of SageOx.

To Dr. Paul Gauthier, for building Aider—a crucial precursor—and walking us through that journey.

To Dr. Rupak Majumdar, for jumping on a flight from Germany the moment he heard about the coding speeds we were hitting ("you folks are treating code like assembler!") and spending the first week of December 2025 getting Claude-pilled with us.

To the generous Seattle-based partners enabling this fledgling startup to make early progress: Foundations and AI House for opening their doors and communities, and to Amazon AWS for their cloud credits.