@usenaive-sdk/vetta-cli

The vetta command line — drive the whole Vetta API from a terminal or a script: agents, sessions, computers, skills, files, identities, vaults, comms, deployments, webhooks, the model proxy, the monthly plan and credits.

It is built on @usenaive-sdk/vetta, the same typed client the dashboard uses, so the CLI and the API never drift apart.

Install

npm i -g @usenaive-sdk/vetta-cli

First run

vetta login --api-key sk_…      # or: vetta login --email … --password …
vetta whoami
vetta plan show                 # an org with no plan is refused everywhere else

Credentials are stored on a named profile, so several organizations can be driven from one machine with --profile.

Shape

vetta <group> <command> [flags]

Groups: agent audit auth computer connections credits deploy domain email file identity keys login logout models org phone plan proxy run session skill vault webhook whoami.

vetta <group> --help lists a group's commands.

Global flags: --json (the default) · --human · --profile <name> · --api-key <key> · --org <org_id> · --api-base-url <url>

Example

# Create an agent, run one task, and read what it cost.
vetta agent create \
  --name summariser \
  --model openai/gpt-oss-120b \
  --system "Answer in one sentence." \
  --budget-usd 2.00 --max-task-usd 0.10 --budget-period day

vetta session run --agent agt_… --message "What is 17 times 3?"
vetta session get ses_… --human

# Let one agent read a short list of trusted hosts, and nothing else on the open web.
vetta agent tools summariser --tool web_fetch --allow-domain docs.example.com --max-content-tokens 2000

# Point an existing Messages-format client at the same models, and check it answers.
vetta proxy url --human
vetta proxy message --model openai/gpt-oss-120b --text "Name three sorting algorithms."

Output

--json is the default because the CLI is meant to be piped. Every command writes a single JSON document to stdout and nothing else; diagnostics go to stderr. --human switches to a readable table when you are driving it yourself.

Exit codes follow the API's error taxonomy, so a script can branch on the failure rather than parsing a message.

License

MIT