Noma Security Pricing: Scope a Quote Without Inventing a Rate

Noma Security presents four products on one platform page: AI-SPM, Access Control, AI-DR, and AI Red Teaming. The same page separates employee endpoints and cloud-built systems into distinct environments. SaaS agent platforms have their own environment. The public site routes a buyer to a contact form rather than a rate card. Procurement needs a scope matrix before comparing the quote with another AI security proposal, because a single annual total can hide product scope and asset counts plus commercial terms for services and meters.

TL;DR

• Noma publishes no public dollar rate or plan allowance on the two pages reviewed on 7 November 2026. It also publishes no commercial meter there.
• The platform names AI-SPM, Access Control, AI-DR, and AI Red Teaming. A quote should mark each one included/optional/excluded.
• Normalize asset classes plus endpoint and cloud surfaces. List SaaS surfaces separately, then normalize services and the meter before comparing annual totals.
• Keep DeepInspect on a separately routed authenticated HTTP model leg. It does not replace Noma's wider asset and endpoint scope or its red-team scope.

Public Noma Security pricing is quote-led

The Noma platform page names the product set and sends visitors to a demo. The Noma contact page asks for a name and work email. It also asks for a phone number, company name, and a description of the request. Neither page published a currency amount or plan tier when I reviewed them on 7 November 2026. They also published no included allowance, overage rate, minimum term, or discount schedule.

The pages provide no basis for an estimated per-agent, per-asset, per-request, or per-user price. I would send back any comparison sheet that fills those cells with a guessed number. A blank cell labelled "vendor to confirm" is more useful than a precise fiction.

The physical artifact I want in the procurement room is one sheet of paper with eight scope rows and three status marks: included/optional/excluded. Put the quoted annual amount at the bottom only after every row has an owner.

Four products need four entitlement decisions

Noma's platform page presents AI-SPM, Access Control, AI-DR, and AI Red Teaming as four products that share intelligence. A commercial proposal should preserve those boundaries instead of reducing them to one line called "AI security platform." Shared intelligence describes product interaction. It says nothing about contractual entitlement.

AI-SPM needs a written inventory scope and a definition of the assets counted. For AI-DR, identify monitored event sources, alert handling, retention, and any response service. Agent access control requires named enforcement locations, supported identities, policy objects, and covered actions. The red-team line should state target count, test frequency, fixture ownership, retest policy, and deliverable format.

For each product, ask for included capacity and optional modules. Ask for technical prerequisites and data retention. Ask for support level and renewal treatment. The AI security vendor evaluation criteria provide the wider route, identity, action, and evidence questions. This narrower matrix has one job: make four named product entitlements commercially legible.

The quote-scope matrix has eight rows

Use the following bullet matrix in the request for proposal. Each answer belongs in the signed order form or an incorporated scope document.

• AI-SPM: List discovery sources and inventory objects. Add assessment frequency and risk records. Include reporting and included capacity. State the counting unit.
• AI-DR: List telemetry sources and detection coverage. Add response actions and retention. Include case workflow and any analyst service. Identify volume assumptions.
• Agent access control: Name supported agent environments and identity inputs. Add policy locations and action points. Include log exports. Define the licensed object or event.
• AI red teaming: Record target applications or agents and test cadence. Add test depth and deliverables. Include retests and remediation support. State which work is automated and which requires people.
• Asset class: Separate models and agents. Separate applications and MCP servers. Separate tools and packages. Separate data connections and other inventoried objects where applicable. Prevent one generic "asset" count from hiding eight populations.
• Deployment surface: Match the three environments on Noma's platform page. Give endpoint agents one entry and cloud-built systems another. Give SaaS agent platforms a third. Add regions and customer-hosted components.
• Services: Split onboarding and architecture work. Split connector setup and policy design. Split red-team execution and training. Split support and incident assistance. Attach hours or service levels where sold.
• Commercial meter: Require the exact billing unit and included quantity. Require the overage rule and minimum commitment. Require the term and renewal method. State the treatment of development or test activity.

The matrix repeats counting fields because "agent" and "asset" may each cover several billable events unless the quote defines one unit with an example. The same applies to "test."

Asset and deployment scope drive comparability

The Noma page describes three environments. Endpoint agents run on employee devices. Cloud agents include AI applications built on services such as AWS Bedrock and Azure AI Foundry. Databricks is another named service, alongside code frameworks. SaaS agent platforms are the third surface. A buyer with 2,000 laptops and 14 cloud agent services also has three SaaS agent platforms. These are three different operational populations. A single asset total strips out that shape.

Ask the vendor to map each included product to each included surface. Assign connectors and credentials to an owner, then identify who operates network paths and upgrades. Evidence export needs its own owner. Record production and staging separately, with development listed when its commercial treatment differs. Put each unresolved term in the quote request.

The AI agent security vendors guide helps place inventory and runtime controls into separate buying categories. It also separates request enforcement. That distinction prevents a broad platform label from turning unlike scopes into an apparent price comparison.

Normalize the commercial meter before annual cost

A comparable proposal defines one billed unit and shows the arithmetic. Possible units in this market include agents and discovered assets. Other possibilities include protected applications and monitored events. Tests and users can also be units, as can a flat platform term. Noma's two public pages do not identify which of these applies. Treat every possibility as a question, never as an assumption.

Request a worked invoice for your expected estate across the three deployment surfaces and eight or more relevant asset classes. Add non-production activity, red-team retests, telemetry retention, and support. Then request the same example at a credible high case. The vendor should show included quantity, overage handling, unit changes, and renewal. Separate software charges from professional services and customer-operated infrastructure.

The best AI gateway guide explains a related cost boundary: model-provider calls and gateway enforcement events can track the same application interaction while remaining separate commercial services. Preserve that separation when Noma and another control appear in one architecture.

DeepInspect

DeepInspect evaluates authenticated HTTP model traffic that an application deliberately routes through its stateless proxy. It can bind application-supplied identity and role to policy for a named data class and model destination. It then commits a per-decision audit record outside the calling application's write path.

That is one separately routed model leg inside a wider Noma program. Noma's AI-SPM and AI-DR remain separate scopes. Its endpoint coverage and SaaS-agent coverage also remain separate, as does its red-team work. Price DeepInspect against the routed model traffic and evidence requirements that it owns. Keep every other Noma asset and service in the quote-scope matrix. Book a technical deep dive at deepinspect.ai.