DECIDE

Your agent is about to act on memory.
Sgraal decides if that action is safe.

Irreversible decisions. Unverified memory. Zero visibility. Until now.

Before every memory-based action, get a risk score, a decision, and a full explanation. Typically 1–2 seconds for full scoring.

Get API Key arrow_forward

Without Sgraal

→ Agent acts on memory from 54 days ago — unverified, unmonitored
→ Irreversible financial decision executes
→ Wrong outcome detected weeks later
→ No trace, no rollback, no explanation

With Sgraal

→ Risk detected before execution
→ Unsafe action BLOCKED automatically
→ Exact entry causing the risk identified
→ Repair plan provided in the same response

Send your memory state and action type

POST /v1/preflight · memory_state + action_type + domain

83 analytics modules evaluate in parallel

Freshness decay (Weibull) · drift detection · provenance · conflict · causal graph · commercial intent.

Get a decision + full explanation

USE_MEMORY / WARN / ASK_USER / BLOCK · Entry Shapley attribution · repair plan.

security

Risk Decision Engine

omega_mem_final: 0–100 risk score.

USE_MEMORY WARN ASK_USER BLOCK

BLOCK = execution never happens

Action multiplier: informational 0.5× · irreversible 1.5× · destructive 2.5×. Assurance score + confidence intervals.

Stop dangerous actions before they happen — not after.

data_exploration

Full Explanation

Entry Shapley: which exact memory entry causes the risk. Causal graph: which entry caused drift.

Know exactly which memory to fix. Debug in minutes, not hours.

EN | DE | FR

build

Repair Plan

REFRESH / DELETE / VERIFY / KEEP / WAIT per entry. Heal + re-preflight in one call.

Don't just detect problems. Resolve them in the same API call.

MTTR prediction: p95 convergence steps.

history

Time Machine

Restore to any previous validated state. Shadow execution with Decision Twins.

No blind debugging. Roll back to any validated state instantly.

verified_user

AI Advertising Integrity Coming Soon

commercial_intent: 0.88

sponsorship_prob: 0.94

AD_INTEGRITY: BLOCK

Your demo key: sg_demo_playground

10,000 free decisions/month · No signup needed

Python SDK

Integrate Sgraal directly into your Python agent workflows.

pip install sgraal

from sgraal import SgraalClient
client = SgraalClient(api_key="sg_demo_playground")
result = client.preflight(
    memory_state=[{"id": "mem_001", "content": "User communication preference",
        "type": "tool_state", "timestamp_age_days": 54, "source_trust": 0.6}],
    action_type="irreversible",
    domain="fintech"
)
print(result.recommended_action)  # BLOCK
print(result.omega_mem_final)     # 78.4

Node.js SDK

Use Sgraal in your Node.js or TypeScript agent projects.

npm install @sgraal/sdk

const { SgraalClient } = require('@sgraal/sdk');
const client = new SgraalClient({ apiKey: 'sg_demo_playground' });
const result = await client.preflight({
  memory_state: [{ id: 'mem_001', type: 'tool_state',
    timestamp_age_days: 54, source_trust: 0.6 }],
  action_type: 'irreversible',
  domain: 'fintech'
});
console.log(result.recommended_action); // BLOCK

REST API

Call Sgraal directly via HTTPS from any language or platform.

curl -X POST https://api.sgraal.com/v1/preflight \
  -H "Authorization: Bearer sg_demo_playground" \
  -H "Content-Type: application/json" \
  -d '{
    "memory_state": [{"id": "mem_001", "content": "User communication preference",
      "type": "tool_state", "timestamp_age_days": 54, "source_trust": 0.6}],
    "action_type": "irreversible",
    "domain": "fintech"
  }'

trending_up

Know Before It Happens

Sgraal doesn't just tell you when memory is unsafe — it tells you when it will become unsafe. Using multiple independent forecasting models, every preflight response now includes days_until_block: the number of days until this agent's memory crosses the BLOCK threshold.

Not “your agent is at risk.” Your agent will need intervention by Thursday.

schedule Preflight response

{
  "omega_mem_final": 38.2,
  "recommended_action": "WARN",
  "days_until_block": 4.2,
  "days_until_block_confidence": 0.73,
  "confidence_calibration": {
    "state": "CALIBRATED"
  }
}

Combines Ornstein-Uhlenbeck, Cox hazard, Kalman trend, and BOCPD changepoint detection.

Ready to make safer decisions?

Get API Key arrow_forward arrow_back

Your agent is about to act on memory.Sgraal decides if that action is safe.