Use the hosted control plane first. No full local runtime required for this path.
NORNR
Governed agent spend before money moves.
Start with governed runtime or browser checkout governance. Name one owner, one counterparty scope and one default pack first. Add MCP control server only after the first scope-bound proof path is live. The lane matters more than the raw SDK surface.
The point is the decision before spend moves, not just the payment afterward.
Observe -> install -> prove
The first NORNR movement should feel calm.
Discovery should show one exposed path. Install should add one review boundary. Proof should show the same decision and packet right away. If a team has to invent the next step, the quickstart is still too abstract.
Name the first consequential path
Use Governance Audit, Shadow AI Discovery or the browser demo when the team already has code or behavior in motion.
Add one calm boundary
Budget Shield, one wrapper patch, one MCP starter or one browser release path is enough for the first install.
Show the defended record immediately
The same first lane should already produce a buyer-safe packet, review reason and export path without a second explanation thread.
What you get
- Choose one real spend lane inside the NORNR control plane
- Run one governed action against hosted policy, approval and proof flows
- See the decision before any real spend moves
- Leave with a buyer-safe proof trail a finance or risk lead can inspect without reading SDK docs
No full local stack required
These snippets use the hosted control plane at https://nornr.com. Start here, then connect an existing workspace, MCP client or finance workflow later.
Pick one first lane only
- Governed runtime: prove one canonical controlled action
- Browser checkout governance: hold risky click-through before checkout lands
- Next package: MCP control server for Claude Desktop, Cursor or Agent Zero once one proof path is already live and owned
Make the first lane boring before you widen it
- One lane
- One owner
- One approved-for-this-scope counterparty posture
- One packet family finance can read later
Open one local tools lane
When one governed runtime or browser lane already makes sense, use the MCP control server to carry the same review and finance trail into Claude Desktop, Cursor or another local agent surface.
Buyer path
This quickstart works best for teams that already have one workflow close to real spend.
If you already have vendor actions, browser checkout or paid tool usage in flight, the fastest next step is to compare one owned lane against the canonical proof path.
Teams with one live lane
Use this quickstart when you can point to one spend lane that needs policy and review now, not just later.
Open the canonical proof path
Compare your workflow against NORNR’s buyer proof before you widen the rollout surface.
One-lane installer
Choose the first lane and let NORNR hand back the whole control shape.
Quickstart should behave like a control-plane installer, not a generic SDK demo. Choose one lane, one surface, one review owner and one scope-bound vendor fit, then keep the proof and finance path attached from the beginning.
Installable default
Install and share
Once the first lane makes sense, use starters and free tools to keep rollout friction low.
The quickest way to widen from hosted proof into a real team workflow is an installable default plus one public-safe artifact surface.
Move from the planner into a real starter
nornr init now opens generic runtime, OpenAI Agents, LangGraph, MCP and browser lanes with one scaffold and one default posture.
Test the outputs buyers will ask for before rollout widens
Use the policy simulator, trust manifest verifier and finance packet previewer to make the first lane more legible before a wider rollout.
Import a named starting posture instead of writing rules from scratch
Community policy packs shorten the path from first lane to a buyer-safe rollout because the review and threshold posture already has a legible starting point.
Turn agent logs into one concrete NORNR fix path
Use the post-mortem tool when you already have a loop, overspend or bad tool path and need to show exactly what policy, packet and review path were missing.
Catch the moment a buyer realizes raw infra is not enough
Use the comparison pages when the team already understands rails, wallet permissions, raw MCP tools or ad hoc approvals and needs the missing layer stated clearly.
SDK quickstart
Install, pick one wedge, run one governed action.
Pick one language, start with one canonical spend lane, and confirm NORNR returns a decision before money moves.
01 Install the SDK package
Start with the hosted path so you can validate behavior without a full local stack.
02 Choose one canonical spend lane
Start with governed runtime or browser checkout governance. Do not try to learn every surface at once.
03 Inspect the returned control-plane decision
Look for approved, queued, or blocked, then open the same trail in the control room and reuse it as the first buyer-facing proof path.
npm install @nornr/sdk
import { Wallet } from "@nornr/sdk";
const wallet = await Wallet.create({
owner: "research-agent",
dailyLimit: 50,
requireApprovalAbove: 20,
baseUrl: "https://nornr.com",
});
const decision = await wallet.pay({
amount: 12.5,
to: "openai",
purpose: "model inference",
});
console.log(decision.status, decision);pip install nornr-agentpay
from agentpay import NornrRuntime
runtime = NornrRuntime.create(
owner="research-agent",
daily_limit=50,
require_approval_above=20,
base_url="https://nornr.com",
)
record = runtime.execute(
action_name="quickstart-model-call",
amount=12.50,
to="openai",
counterparty="openai",
purpose="model inference",
callback=lambda: {"ok": True},
)
print(record.execution_status)
if record.requires_handoff:
print("Review in NORNR:", record.decision.approval_url if record.decision else None)
elif record.execution_status == "completed":
print("Intent cleared before spend moved.")
else:
print("Blocked:", record.to_summary_dict())How to read the result
- approved: policy allows the spend immediately
- queued: control routes the intent into review before spend can move
- blocked: policy blocked the spend
Where to go after first success
- Connect an existing workspace with Wallet.connect(...)
- Open the control room to inspect approvals, anomalies and evidence
- Use the playground to walk through replay viewer and A2A escrow flows before wiring a live workspace
- Move into wrappers, browser guards or MCP once the first lane is working
- Expand only after the first decision path is defensible for a buyer or operator