Norla
⌘K
Bag Orders

THE NORLA PLATFORM

Software for multi-agent
marketing systems.

Norla provides online, non-downloadable software for implementing multi-agent systems: several intelligent agents that interact to carry out marketing work, with the handoffs and human approvals written down. It is used in the browser for the term of your plan, alongside the Extensions store, the Academy and the operator network.

Open Norla Orchestrator
01

Agent Hosting

Deploy, monitor, and scale production agents with versioned configuration and clear ownership.

Learn more
02

Multi-model Routing

Route every task by quality, latency, cost, context, and fallback requirements.

Learn more
03

Extensions store

Agent templates by Norla, plus resources and services from independent developers, installed into Orchestrator workspaces.

Learn more
04

Marketing Operations

Coordinate content, campaigns, lifecycle, creative, reporting, and approvals.

Learn more
05

Enterprise Controls

SSO, roles, audit history, data boundaries, and visible human decision points.

Learn more
06

Evaluation & Analytics

Measure task success, operating quality, cost, latency, and business outcomes.

Learn more

ORCHESTRATOR PLANS

Temporary use, by subscription.
Nothing to download.

Trial$014 days

14 days of Orchestrator in the browser, once per account

  • 3 workspaces, 8 agents each
  • 60 runs, sandbox engine
  • Handoff contracts and approvals
  • Run log and history
Start the trial
Plus$29/mo

One operator building a first system or two.

  • Up to 3 workspaces
  • Up to 4 agents per workspace
  • 60 runs per month
  • Run history kept 30 days
  • Handoff contracts and approvals
Subscribe to Plus
Pro$79/mo

One operator running several systems in production.

  • Up to 20 workspaces
  • Up to 8 agents per workspace
  • 400 runs per month
  • Run history kept 90 days
  • Model engine when a provider is connected
  • Priority in the course clinic
Subscribe to Pro
Max$229/mo

A team's worth of systems on one account.

  • Up to 60 workspaces
  • Up to 12 agents per workspace
  • 2,000 runs per month
  • Run history kept 1 year
  • Model engine when a provider is connected
  • Priority in the course clinic
Subscribe to Max

NORLA / WORKING GUIDE

Evaluate a workflow before expanding its reach.

A bounded test gives an AI system a job you can inspect, compare and stop.
01

Define the operating boundary

Write one input-to-output sentence and name the person who reviews the result. Separate preparing an action from executing it, and identify which external systems the workflow is allowed to touch.

Document missing inputs and stop conditions before testing. A system that asks for clarification can be behaving correctly; an attractive completed answer is not automatically a valid result.

Read the job-boundary discussion →
02

Use a representative comparison

Build a small set of ordinary, incomplete, conflicting and out-of-scope inputs. Keep the expected behavior visible so a reviewer is not inventing the acceptance criteria after seeing the answer.

Compare versions on the same cases. Record the concrete failure, its consequence and whether a human review step can contain it. Do not turn an arbitrary success percentage into a production guarantee.

Study practical evaluation →
03

Plan the handoff into operation

Name the owner of access, cost monitoring, error handling and version changes. Keep the previous configuration and an explicit reason for returning to it.

The catalogue describes workflow scopes and learning resources. Confirm actual hosting, integrations, support and usage costs separately before treating a system as provisioned.

Compare system scopes →

NORLA EDITORIAL / FIELD NOTES

The operator’s reading room.

Working methods, decisions to document, and useful questions to take into your next project.

All field notes