Getting started.

1. Install the SDK

The Raven SDK ships for TypeScript and Python. Pick whichever fits your environment.

npm install @raven/sdk
# or
pnpm add @raven/sdk

pip install raven-sdk

2. Authenticate

Raven uses short-lived bearer tokens issued per workspace. Generate a token from the workspace settings panel and store it as an environment variable.

export RAVEN_TOKEN=rvn_live_xxxxxxxxxxxxx
export RAVEN_WORKSPACE=acme-prod

Token scopes

• muninn.read — read observations from Muninn
• muninn.write — ingest into Muninn
• huginn.read — read evidence and decisions from Huginn
• field.execute — trigger workflows in Sleipnir

3. Run your first query

The minimal end-to-end is: create a client, scope it to a namespace, and ask it a question. The example below queries Muninn for recent observations of a topic and prints the top results.

import { Raven } from "@raven/sdk";

const raven = new Raven({
  token: process.env.RAVEN_TOKEN!,
  workspace: process.env.RAVEN_WORKSPACE!,
});

const results = await raven.muninn.search({
  query: "anomalies in mainline pump telemetry",
  since: "24h",
  limit: 10,
});

for (const r of results) {
  console.log(r.score, r.subject, r.source.uri);
}

Every record carries provenance — the source URI, ingestion time, and a chain-of-custody hash. That metadata is yours to surface in your own downstream UIs.

Next steps

• Read the architecture overview to understand how the pieces fit together.
• Explore the SDK reference for every available operation.
• Review the security model before promoting from sandbox to production.