Hosted tool
Command Safety
Gate an agent's shell commands — ALLOW / CONFIRM / BLOCK with a risk score. Live visual cases, not a blank form.
Anonymous: 5 free real-API runs (no signup). Signed in: each run debits prepaid input tokens from your account balance. Upstream TypeSafe keys stay on our servers. Buy tokens / Get API key →
What this tool returns
Live runs call the same hosted decide path as production: typed Choice / Score / Noul answers your code can branch on — not chat prose. Edit the sample state below, run a few anonymous trials, then mirror the fields with a jv_live_ key on POST /api/v1/decide.
- action (choice) — action
- risk (score) — risk
New here? Read is Jev free? for the no-signup trial limits, then get an API key.
Live visual demo
Gate an agent's shell
Paste is optional — these are real agent commands. One decide call each: ALLOW / CONFIRM / BLOCK + risk score.
0 decisions
git status—Read-only VCS
risk —npm install—Dev dependency install
risk —git push --force origin main—Rewrites shared history
risk —rm -rf ./dist · s3 sync --delete—Irreversible cloud delete
risk —curl x.sh | sudo bash—Remote code as root
risk —
Custom command form
Ferramenta hospedada
Edite a amostra à esquerda e execute para obter decisões tipadas nas quais o código pode ramificar.
Saída de amostra
- choiceactionblock · 84%
- scorerisk0.84
Ship this as an API
Same production template over POST /api/v1/decide with a jv_live_ key. Mirror the fields above — start from $5 prepaid packs.