Kernel Docs
Quickstart
Get Kernel running locally and send your first evaluation in under 10 minutes.
1. Run Kernel locally
Download the Evaluation Kit — a self-contained Docker stack with API, dashboard, and Postgres.
# Download and start
curl -LO https://kerneva.com/evaluation-kit.zip
unzip evaluation-kit.zip
cd evaluation-kit
cp .env.example .env
docker compose up
curl -LO https://kerneva.com/evaluation-kit.zip
unzip evaluation-kit.zip
cd evaluation-kit
cp .env.example .env
docker compose up
Open http://localhost:3000 to see the dashboard.
2. Send a sample evaluation
./sample/run.sh
This sends 5 escalating refund requests. Kernel flags the trajectory change. Go back to the dashboard to inspect the investigation.
Or use Kernel Hosted
Kernel Cloud is in private preview. The Docker flow below works today. Hosted API and console will be available once domain setup is complete.
Use this path when you want the fastest proof with no local stack.
# 1. Create an account and get an API key
https://console.runtime-trust.com
# 2. Point your client at the hosted API
export RUNTIME_TRUST_API_KEY="sk_live_..."
export RUNTIME_TRUST_API_URL="https://api.runtime-trust.com"
# 3. Send a request
curl -X POST https://api.runtime-trust.com/evaluate \
-H "Authorization: Bearer $RUNTIME_TRUST_API_KEY" \
-H "Content-Type: application/json" \
-d {
"agent_id": "payment-bot",
"session_id": "ses-abc-123",
"action_type": "refund",
"amount": 150.0
}
https://console.runtime-trust.com
# 2. Point your client at the hosted API
export RUNTIME_TRUST_API_KEY="sk_live_..."
export RUNTIME_TRUST_API_URL="https://api.runtime-trust.com"
# 3. Send a request
curl -X POST https://api.runtime-trust.com/evaluate \
-H "Authorization: Bearer $RUNTIME_TRUST_API_KEY" \
-H "Content-Type: application/json" \
-d {
"agent_id": "payment-bot",
"session_id": "ses-abc-123",
"action_type": "refund",
"amount": 150.0
}
Hosted returns the same decision, risk score, trust score, signals, and evaluation ID as the self-hosted path. The only difference is where the runtime runs.
What happens next
Record the execution outcome with the review endpoint, then inspect it in Evaluation History and the incident view.
PUT /evaluations/{evaluation_id}/review{
"status": "resolved",
"resolution": "completed",
"impact": "{"external_id":"pay_abc123","duration_ms":842}"
}
"status": "resolved",
"resolution": "completed",
"impact": "{"external_id":"pay_abc123","duration_ms":842}"
}