Skip to main content
Vetta is the autonomous-company infrastructure: the harness, the managed agent runtime, sessions, budgets, identities and tools. Naive is the layer on top of it — the blueprints and templates a whole company is installed from, the catalogue that publishes them, and the studio you run them in. These pages are the Vetta reference. Every page below documents the infrastructure, which is why the prose, the CLI and the API all say vetta.
Naive is not a second platform. A blueprint has no runtime of its own: everything it installs is an ordinary Vetta resource, created through the same API you could call yourself.

​
Where the line sits

Vetta — the infrastructure

Agents, sessions, the durable runtime, harnesses, the completion window, budgets and the ledger, computers, files, skills, identities and vaults, deployments and webhooks — reached over one REST API, one SDK and one CLI.

Naive — the layer on top

Blueprints and their templates: a company as a repository — its dashboard, its crew, its personas and its timers — plus the published catalogue of those pairs and the studio that installs and operates them.

​
The four surfaces

​
What a blueprint actually installs

A blueprint is a public repository with a naive.config.ts in it — naive template agency clones usenaive/agency-blueprint, naive template media clones usenaive/media-blueprint. A template is data inside that repository: the crew and its prompts, the tool allow-lists, the schedules, and the words its screens print. One repository carries every template its blueprint has. Applying that config — naive up in the clone, or the studio’s install flow — reconciles your organization against it. Every row it writes is a Vetta resource you could have created yourself: The report names each one created, updated, unchanged, deleted or refused. It is idempotent — every resource is keyed by its config name — and nothing is deleted by omission. The full config surface is in the naive CLI reference.

​
Which one am I touching?

  • Standing up or switching a whole company — Naive. Install from the studio, or clone the blueprint and run naive up. Switching template is an edit plus another naive up; the app, its URL and its data stay where they are.
  • Changing an agent a blueprint installed — the resource is Vetta’s, but the config is the source of truth: edit naive.config.ts and re-run naive up, because the next apply converges a field edited by hand back to what the config declares.
  • Anything not from a blueprint — Vetta only. An agent you create by hand, a session, a computer, a key: nothing about blueprints is required to use the platform, and the quickstart never mentions one.
  • Watching work happen — sessions, events, files and cost read the same whether the agent came from a blueprint or from POST /v1/agents.
vetta and naive are two binaries pointing at the same bundle from @usenaive-sdk/vetta-cli; the name selects nothing. naive up is vetta naive up.

Quickstart

Create an agent, run it, and read the meter in under five minutes.

The naive CLI

claim, up, template, catalog, installs — and the whole config surface.