DeAlgoOpen workspace
Core decision API setup. For restricted financial agents, use the commerce quickstart.

Quickstart

First verdict in 60 seconds.

No setup required. Use a sandbox key and make your first call instantly.

Put DeAlgo between your agent and its action.

Before
if agent.decides("send_email"):
    send_email()
After
verdict = csc.decide(
    action_type="send_email", ...
)

if verdict.decision == "APPROVE":
    send_email()
01

Get a sandbox key

sk_test_
Sign in to mint a key

Free. Google or GitHub. Sandbox keys don't count toward any quota.

02

Make the call

Make your first governed decision in 60 seconds. Pick a flavor — curl is fastest. The bearer token below is a placeholder until you mint a key above.

curl -X POST https://dealgo-portal.vercel.app/api/v1/decide \
  -H "Authorization: Bearer dealgo_sk_test_REPLACE_ME" \
  -H "Content-Type: application/json" \
  -d '{
    "action_type": "send_email",
    "risk_score": 0.1,
    "confidence": 0.95,
    "intent": "reply to customer with quote",
    "metadata": { "agent_id": "support-bot-7" }
  }'

Paste this into your terminal. You'll get a real decision immediately.

What comes back

Same shape regardless of language. decision is the only field your code branches on.

{
  "decision":             "APPROVE",
  "action":               "EXECUTE",
  "reason":               "Action approved.",
  "reason_raw":           "permitted",
  "retry_after_ms":       null,
  "stimulus_id":          "ec9262c8-79e1-4665-...",
  "latency_ms":           41,
  "environment":          "test",
  "idempotency_replayed": false,
  "decision_log_id":      "cmoi2g2yzze09y0xww82kb6z8",
  "v2_bias":              0
}
View full decision trace→ /decisions/<decision_log_id>

See exactly why your action was approved or blocked.

  • APPROVE → execute the action.
  • DELAY → wait retry_after_ms and retry. Use the same Idempotency-Key header to avoid double-metering.
  • DENY → don't retry. Surface reason to the user.

What just happened

That single call landed three things at once:

  • A verdict — your agent now knows whether to execute, delay, or stop. See it on your dashboard.
  • A trace — each step the runtime evaluated, with the reason it produced, in the order it happened. Open any row in /decisions to inspect the full decision trace.
  • A verifiable proof — the row is hash-chained into your workspace's tamper-evident audit log. A third party can recompute every hash without trusting our database. SSI Protocol™ →