Typed memory · visible state · human gates
Give your agents a world they can understand.
Vivary is a standard and scaffolder for agent-native workspaces. Start clean or safely adopt the project you already have—without locking your work into a platform.
npm create @vivary@latest my-workspace- Plain Markdown + YAML
- Claude Code + Codex
- MIT licensed
One world. Four layers.
A small atmosphere for long-running work.
- The substrate
tropo
A typed knowledge graph for what the workspace knows, where truth lives, and what depends on it.
- The operating layer
strato
Visible state, reusable skills, private boundaries, and a loop that learns without bloating active context.
- The protective filter
ozone
Graph-aware review and human gates for changes that are durable, destructive, or outward-facing.
- The coordination layer
exo
Claims, conflicts, and role contracts for the moment one agent becomes many.
The product thesis
Agents do better work when the workspace has a shape.
Flat memory files turn into junk drawers. Giant instruction files consume the context they were supposed to protect. Vivary gives the project a typed, navigable structure with one visible state surface and explicit gates at the edges.
No greenfield required
Point Vivary at the mess you already have.
The adoption path is deliberately read-first, additive, and gated. You can feel the value before giving Vivary permission to write anything.
- 01
Map the terrain
tropo map --root .Read a large repo, vault, or docs tree without opening everything. No config and no writes.
- 02
Adopt without wrecking it
create-vivary adopt .Preview the additive workspace shell first. Existing files remain byte-identical.
- 03
Keep it healthy
create-vivary doctor . --trendSee graph, routing, and context-budget drift before the workspace quietly rots.
Proof before promise
The claim is measurable.
Vivary should reduce the tokens and wrong-file opens an agent spends finding project truth. The roadmap prioritizes a rerunnable benchmark, brownfield proof, retention through doctor, and honest public signals.
Quick answers
Before you bring it into a real project.
The short version of what Vivary is, what it touches, and where the current evidence stops.
What is Vivary?
Vivary is a plain-file standard and scaffolder for agent-native workspaces. It gives agents typed project memory, visible state, reusable skills, private boundaries, and explicit human gates.
Can I use it with an existing project?
Yes. The brownfield path starts read-only with tropo map, previews an additive workspace shell, and leaves existing files byte-identical until you approve a write.
Is Vivary an AI agent or model?
No. Vivary is the workspace structure around your agents. It works with tools such as Claude Code and Codex without replacing the agent, editor, model, Git, or issue tracker you already use.
Does it upload or index my files?
Not by default. The core is local and dependency-light. Networked providers, semantic indexing, daemons, and MCP integrations are optional edges that require explicit installation and approval.
Am I locked into Vivary?
No. Project truth stays in Markdown, YAML, and ordinary folders. You can inspect it, version it with Git, edit it anywhere, and remove the Vivary tooling without trapping your work in a hosted platform.
Is it open source and production-ready?
Vivary is MIT licensed and its core packages are public. The architecture and basic adoption path work today; benchmark results, broader brownfield proof, and retention evidence are still active roadmap work.
The deeper argument
Why agent workspaces need a standard.
The white paper defines the system invariants, architecture, threat model, human-gate protocol, evidence ledger, and benchmark that keep Vivary's claims falsifiable.