create-vivary 0.2.1

Scaffold a complete Vivary agent workspace — the create-t3-app for agent-native workspaces.

create-vivary

Scaffold a complete Vivary agent workspace: tropo config, strato workspace files, runtime skills, private-memory boundaries, progressive module indexes, and a starter typed graph.

Install & scaffold

pip install create-vivary                     # or run without installing: uvx create-vivary …
create-vivary my-workspace --preset coding    # interactive wizard on a TTY
create-vivary my-codebase --preset coding --active-context cocoindex-code
create-vivary doctor my-workspace

# Agent-mode (no prompts, machine-readable output):
create-vivary init . --preset coding --auto --size large --yes --json

# Reconfigure storage on an existing workspace:
create-vivary wizard my-workspace --auto --storage embedded --yes --json

create-vivary <name> is shorthand for create-vivary init <name>; pass init / doctor / wizard explicitly whenever you prefer. The same UX is available on npm via the @vivary/create launcher (npm create @vivary my-workspace), versioned in lockstep.

Local use

python packages/create-vivary/create_vivary.py init sandboxes/coding-demo --preset coding
python packages/create-vivary/create_vivary.py doctor sandboxes/coding-demo
python packages/tropo/tropo.py check --root sandboxes/coding-demo

Presets share the same agent OS shell, then seed a different starter graph. Each starter module is generated as modules/<id>/index.md so agents route through a small module index before opening deeper context:

Preset	Module	First slice	Verification
coding	codebase	local-ci-baseline	local-checks
second-brain	knowledge-base	capture-routine	retrieval-smoke
writing	manuscript-system	draft-review-loop	editorial-review

The command is local-only. With --storage embedded or --auto on a large workspace, it self-installs vivary-tropo[embedded] (LanceDB) with a confirmation prompt, or silently with --yes. Use --dry-run to simulate without writing anything. Use --no-wizard to skip interactive prompts entirely.

For coding workspaces, --active-context cocoindex-code adds an optional CocoIndex-code sidecar profile: active-context skills for Claude/Codex-style agents, local policy docs, graph nodes, and .cocoindex_code/ in .gitignore. It does not auto-install CocoIndex-code, create an index, or enable MCP; the generated docs give the approved ccc init / ccc index path.

doctor validates the generated shell, privacy ignores, module directory indexes, and typed graph:

python packages/create-vivary/create_vivary.py doctor sandboxes/coding-demo --json

---

Website & docs: https://vivary.vercel.app/