Quickstart

From zero to your first RCAN message in 5 minutes.

01Install the SDK

# Install

pip install rcan

# Optional extras

pip install rcan[http]    # registry client (httpx)

pip install rcan[crypto]  # Ed25519 message signing

pip install rcan[all]     # everything

# Install

npm install @continuonai/rcan-ts

# or yarn

yarn add @continuonai/rcan-ts

# Import

import { RobotURI, RCANMessage, ConfidenceGate } from '@continuonai/rcan-ts';

# CDN (no build step)

<script src="https://unpkg.com/@continuonai/rcan-ts/dist/rcan.iife.js"></script>

<script>const uri = RCAN.RobotURI.parse('rcan://registry.rcan.dev/acme/arm/v1/unit-001');</script>

# Registry API (curl)

curl https://rcan.dev/api/v1/robots \

  -H 'Content-Type: application/json' \

  -d '{"manufacturer":"acme","model":"arm","version":"v1","device_id":"unit-001"}'

02Build a Robot URI

Every robot has a globally unique, resolvable address. No more "which unit was it?"

address.py

from rcan import RobotURI

uri = RobotURI.build(
    manufacturer="acme",
    model="robotarm",
    version="v2",
    device_id="unit-001",
)
print(uri)
# rcan://registry.rcan.dev/acme/robotarm/v2/unit-001

# Parse from a string
uri = RobotURI.parse("rcan://registry.rcan.dev/acme/robotarm/v2/unit-001")
print(uri.namespace)      # acme/robotarm
print(uri.registry_url)   # https://registry.rcan.dev/registry/...

03Gate on AI Confidence

The gate doesn't trust the model — that's the point. The model is right 99.99% of the time. The gate catches the 0.01%.

gate.py

from rcan import RCANMessage, ConfidenceGate

gate = ConfidenceGate(threshold=0.8)
confidence = 0.91  # from your AI model

if gate.allows(confidence):
    msg = RCANMessage(
        cmd="move_forward",
        target=uri,
        params={"distance_m": 1.0},
        confidence=confidence,
        model_identity="Qwen2.5-7B-Q4",
    )
    print(msg.to_json(indent=2))
else:
    print(f"Blocked: confidence {confidence} < threshold {gate.threshold}")

For high-stakes actions, add a HiTLGate to require explicit human approval. See §16 AI Accountability.

04Create a Tamper-Evident Audit Record

Every safety-critical action can be sealed into an HMAC-chained record — forensic-grade proof of what the system did, when, at what confidence.

audit.py

from rcan import CommitmentRecord
from rcan.audit import AuditChain

chain = AuditChain(secret="your-hmac-secret")

record = chain.append(CommitmentRecord(
    action="move_forward",
    robot_uri=str(uri),
    confidence=0.91,
    model_identity="Qwen2.5-7B-Q4",
    params={"distance_m": 1.0},
    safety_approved=True,
))

print(f"Sealed: {record.content_hash[:12]}...")
print(f"Chain valid: {chain.verify_all()}")  # True

# Export JSONL for long-term storage or audit
with open("audit.jsonl", "w") as f:
    f.write(chain.to_jsonl())

05Register Your Robot

Get a globally unique RRN (Robot Registry Number). Free.

Terminal

# If using OpenCastor — built into the setup wizard
castor wizard
# → Step 13: Register with rcan.dev? [Y/n]  →  ✅ RRN-00000042

# Or standalone
castor register --config myrobot.rcan.yaml

06Validate Your Config

Terminal

# Check L1/L2/L3 conformance
rcan-validate config myrobot.rcan.yaml

# Expected output:
# ✅ L1 — Addressing + message format: passed
# ✅ L2 — Auth + confidence gates: passed
# ⚠️  L3 — hitl_gates not configured (§16)
# Result: L2 (1 warning)

# Or with OpenCastor
castor compliance --config myrobot.rcan.yaml

What's next

Protocol Spec

Full RCAN v1.2 specification

Conformance Tests

L1/L2/L3 test suite

rcan-py on GitHub

Python SDK source + issues

rcan-ts on GitHub

TypeScript SDK source + issues

OpenCastor

Robot runtime with RCAN built in

Get your free RRN