Vetta: Long-Horizon Agents

Only pay while it's working. #

It saves progress and sleeps between steps, so an eight-hour task doesn't bill you for eight hours.

Naïve raises $28.5M Series A to build autonomous company infrastructure

Vetta is a prod-ready harness that runs in managed infrastructure. Best for cost efficiency in long-running tasks.

Tasks finished (%)

Cost ($ / finished)

Loose

$0.171

Priority

$0.233

Immediate

$0.298

Claude Code

$0.872

Hermes

$1.095

Pi

$1.102

Codex

$1.325

It saves progress and sleeps between steps, so an eight-hour task doesn't bill you for eight hours.

Name any model and how hard it should think, then pick a speed — work that isn't urgent runs about 40% cheaper.

Latency

answers soonCost / completed task

$0.2332It gets a shell, a real browser, your files, 10,000+ connected apps, and any MCP server you point it at.

A spend cap the run can't blow past, and a list of actions it has to ask a human about first.

A list of roles, and the lead agent hands each one a single task instead of the whole chat history.

A disposable Linux box with a shell, a filesystem, and a real browser that s when the agent sleeps.

Built for production

A typed client, a CLI, and a REST endpoint — the same operation in all three. The spend cap is part of the config, so every agent has a ceiling from the moment it exists.

import Vetta from "@usenaive-sdk/vetta";

const vetta = new Vetta();

const agent = await vetta.agents.create({
  name: "nightly-triage",
  model: "zai-org/GLM-5.2-FP8",
  system: "You are a concise triage agent.",
  window: "loose",
  skills: ["triage-policy"],
  budget: {
    cap_micro_usd: 50_000_000,
    max_task_micro_usd: 5_000_000,
    period: "month",
  },
});