Developers / Quickstart

Three paths. One authority runtime.

Start with the open source path if you want to self-host the digital flow. Request platform access if you need enterprise-grade intent verification with Authority Tokens, audit, and compliance crosswalks. Or jump straight to physical AI if you're integrating cobots, AGVs, drones, or surgical primitives — same authority runtime, sub-second tokens verified at the edge.

Clone IntendedOps Request platform access Physical AI / Manufacturing Full documentation

Path A — Open source

Clone IntendedOps. Run it locally. Governed operations in five minutes.

IntendedOps is the complete governed back-office operating system. Apache 2.0, self-hosted, and shipped with native governance built in. No API keys. No accounts. No external services required to try it. SQLite + the built-in auth provider let you go from git clone to your first governed intent without provisioning anything.

bash

# Open source path — IntendedOps, Apache 2.0, self-hosted.
git clone https://github.com/intended-so/intendedops.git
cd intendedops
pnpm install

# Minimal config — SQLite, native governance.
cat > intendedops.config.ts <<'EOF'
import { defineConfig } from '@intendedops/config';

export default defineConfig({
  name: 'My SaaS',
  database: { provider: 'sqlite' },
  auth: { provider: 'builtin' },
  governance: { provider: 'native' },
});
EOF

pnpm build && pnpm start

# Open the operator console at http://localhost:3001
# Submit your first governed intent, watch the approval flow,
# inspect the audit chain. All local. All Apache 2.0.

IntendedOps — self-hosted, native governance, no external dependencies.

github.com/intended-so/intendedops →README →

Path B — Intended platform

Request access. Get an API key. Use the platform SDKs.

The Intended platform adds what enterprise deployments need on top of native governance: Large Intent Model classification, Enterprise Capability Engine mapping, RS256-signed Authority Tokens with single-use nonces, hash-chained immutable audit evidence, and compliance crosswalks for SOC 2, NIST AI RMF, and EU AI Act. Platform access is granted to customers with a signed agreement.

Platform SDK example

typescript

// Platform path — commercial Intended runtime.
// Requires platform access. Request at /contact.

import { IntendedSdk } from "@intended/sdk";

const intended = new IntendedSdk({
  apiKey: process.env.INTENDED_API_KEY,
});

const decision = await intended.execute({
  action: "deployment.trigger",
  target: "service-checkout",
  environment: "staging",
  adapter: "github-actions",
});

// decision.outcome: "AUTHORIZED"
// decision.token: "eyJhbGciOiJSUzI1NiIs..."
// decision.riskScore: 18
// decision.auditHash: "sha256:7f3a2b..."

Platform SDKs are distributed to customers. Request access at /contact.

Or call the platform API directly

bash

# Platform path — raw HTTP against the Intended API.
# Requires platform access.

curl -X POST https://api.intended.so/intent \
  -H "Authorization: Bearer $INTENDED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "tenantId": "your-tenant",
    "actor": { "id": "github-actions", "type": "service" },
    "targetSystem": "service-checkout",
    "proposedAction": "deployment.trigger",
    "riskContext": {
      "baseRiskScore": 18,
      "environment": "staging"
    }
  }'

# → { "intentId": "...", "decision": "AUTHORIZED",
#     "token": "eyJhbGci...", "auditHash": "sha256:..." }

The Intended platform API — commercial, customer-facing.

Request platform access Platform SDK catalog API reference

Path C — Physical AI / Manufacturing

Gate cobot, AGV, drone, or surgical actions on Authority Tokens.

Same authority runtime, different shape. Robotics agents emit ROS2 actions, OPC-UA method calls, MAVLink commands — not English. Intended classifies the structured payload into an OIL v2 category, mints a sub-second RS256 token bound to the robot's identity, and verifies it locally at the edge with the Rust verifier crate. Layered above your existing safety PLC, never in place of it.

bash

# Five-minute physical-AI quickstart — no robot required.
# The example uses a JSON snapshot of work-cell state so you
# can run the full loop on a laptop.

git clone https://github.com/intended-so/intended.git
cd intended/examples/physical-ai/pick-and-place

pip install "git+https://github.com/intended-so/intended.git@main\
#subdirectory=packages/intended-python"

export INTENDED_API_KEY=mrt_xxxxxxxx
export INTENDED_API_URL=https://api.intended.so

python agent.py goals/pick-workpiece.json
# GOAL: pick-workpiece workpiece-A4-7 (ros2:action:moveit_msgs/Pick)
#   CLASSIFY oil_code=OIL-1502 confidence=0.97 safety_bit=true
#   ► EXECUTING pick-workpiece workpiece-A4-7
#     token (truncated): eyJhbGciOiJSUzI1NiIs…

python agent.py goals/malformed.json
# ✖ DENIED — falling back to safe-default: stop

REF-P1 — pick-and-place reference cobot, runs on a laptop.

Physical AI overview Physical-AI docs Rust edge verifier

Upgrade path

Start with IntendedOps. Switch to the platform with one config field.

Both paths speak the same governance contract. If you start with IntendedOps native governance and later need platform-grade verification, change one line of config: governance: { provider: 'native' } becomes governance: { provider: 'intended', apiKey: ... }. No rewrites. No migrations. Your intents, policies, and audit shape stay the same — the decisions just flow through the Intended platform instead of the native engine.

See the full open source page →What the Intended platform adds →

Go live fast.

Start Pilot Request Demo

View Pricing