Last updated: November 3, 2026. Testing track: Documentation Reviewed.
Noma discovers AI agents, MCP servers, skills and models across SaaS, cloud and employee endpoints, then governs what each agent may access and blocks risky behavior at runtime. It also offers AI red teaming and AI supply chain scanning. It is aimed at security teams governing agents rather than at controlling employee use of AI websites.
Score: 7.1 / 10.
Initial review November 3, 2026.
This review is based on Noma Security’s public documentation, product pages, pricing pages and press coverage as of November 2026. We have not had lab access or a vendor demo.
Scores use the same published rubric and formula as every review on this site. Sources are listed at the end of the review.
Coverage breadth
Detection accuracy
Deployment friction
Policy & control depth
Framework alignment
Support & documentation
20%
20%
15%
15%
10%
10%
10%
8
7
7
8
6
6
6
SaaS, cloud and endpoint agents, MCP, models and supply chain, but no workforce browser coverage.
Documented session-aware detection of injection, poisoning and drift, with no public benchmarks.
Reuses existing gateways, EDR and MDM, though full coverage still means several integration points.
Per-tool, per-action, identity-scoped controls with human approval and kill switch.
Mentions framework mappings but does not name them on reviewed pages.
Marketplace-only figures, including usage rates.
Detailed product pages and research, but no public docs portal found.
Score
8
Notes
SaaS, cloud and endpoint agents, MCP, models and supply chain, but no workforce browser coverage.
Score
7
Notes
Documented session-aware detection of injection, poisoning and drift, with no public benchmarks.
Score
7
Notes
Reuses existing gateways, EDR and MDM, though full coverage still means several integration points.
Score
8
Notes
Per-tool, per-action, identity-scoped controls with human approval and kill switch.
Score
6
Notes
Mentions framework mappings but does not name them on reviewed pages.
Score
6
Notes
Marketplace-only figures, including usage rates.
Score
6
Notes
Detailed product pages and research, but no public docs portal found.
Granular agent and MCP access control down to individual tools and actions, scoped by IdP identity (Noma Agent Access Control page).
Enforces through existing gateways, agent hooks, EDR, MDM and APIs instead of requiring a new gateway (Noma Agent Access Control page).
Covers agents on employee endpoints as of September 2026, in addition to SaaS and cloud agents (Noma endpoint announcement).
One of the few vendors in this group with AWS Marketplace figures visible, including usage-based rates (AWS Marketplace listings). Well funded, with a $100M Series B in July 2025 and reported Fortune 500 adoption (DRJ coverage of the Series B).
Specific framework mappings are not documented on the pages reviewed.
No public documentation portal or SIEM integration documentation was found.
Marketplace contract figure of $200,000 per year signals an enterprise price point. No public detection benchmarks or false positive data found. Agent Boundaries for endpoints was described as planned, not shipped, in September 2026.
Enterprises with many agents across SaaS, cloud and developer endpoints that want a single registry and least-privilege access control over agents and MCP tools, with runtime blocking and red teaming from one vendor, and that already run gateways, EDR or MDM they want to reuse as enforcement points.
Small and mid-sized organizations that need published, low-cost pricing, or teams whose main goal is stopping employees from pasting sensitive data into AI websites. Workforce AI usage control tools such as Harmonic, Nudge or Prompt Security fit that better; teams wanting a lighter guardrail API for a single LLM app may prefer Lakera.
Not published on vendor site. AWS Marketplace: Noma Security Platform at $200,000 for a 12-month contract (unit undefined); Noma Security Platform for Security Hub Extended is usage-based at $8.00 per million tokens scanned (hybrid), $16.00 per million tokens (SaaS), $130.00 per asset scanned and $650.00 per red teaming test.
See how it compares in our Best AI Agent and MCP Security Platforms 2026 ranking. The poor-fit section above names alternatives for buyers it does not suit, and our methodology explains how every product is scored.
If Noma Security grants lab access, we would run the following scenarios and upgrade this review to Lab Tested.
Discovery of a known set of agents, MCP servers and tool permissions.
Prompt-injection scenarios that try to trigger unsafe tool calls through MCP.
Block, warn, redact and allow behaviors against a configured policy.
What is logged, what is not, and retention behavior.
Microsoft Entra ID and Okta where supported.
Time from access to enforced policy in a representative environment.
Added latency on standard prompt sizes.
Based on public documentation. “Not documented” means we could not find it in public material, not that the product lacks it.
| Capability | Noma Security |
|---|---|
| Inline prompt DLP | Partial. AI-DR masks sensitive data inline in agent traffic; workforce prompt DLP for AI websites not documented |
| Shadow AI discovery | Yes. AI-SPM inventories agents, MCP servers, skills and tools across SaaS, cloud and endpoints |
| Blocking unapproved AI tools | Partial. Blocks agents, MCP servers and skills via access control; blocking AI web apps not documented |
| File upload inspection | Not documented. Model artifact scanning documented, user file upload inspection not |
| Secure AI workspace (sanctioned chat) | Not documented. No vendor-provided chat interface |
| Built-in AI vendor risk scoring | Partial. AI supply chain and asset risk prioritization; vendor risk ratings not documented |
| Microsoft 365 connected AI app review and control | Partial. Covers Microsoft Copilot agents; OAuth connected-app review and revoke not documented |
| Google Workspace connected AI app review and control | Not documented. Gemini listed as an integration logo only |
| Governance scorecard, attestation and board reporting | Partial. Compliance reporting with inventory, test results, audit trails and framework mappings; board reports not documented |
| AI agent and MCP security | Yes. Agent Access Control with per-tool MCP permissions, agent identity and runtime AI-DR |
| Prompt injection defense | Yes. Direct, encoded and indirect injection through tool responses |
| AI red teaming | Yes. AI Red Teaming product with multi-turn attacks; priced per test on AWS |
| SIEM integration | Not documented. No SIEM connector found on pages reviewed |
SaaS or hybrid (per AWS listing). Enforcement through existing control points: agent hooks, AI and MCP gateways, EDR and MDM, and direct APIs; an MCP gateway is not required. Native guardrail in TrueFoundry. Typical rollout time not published.
AI agents on SaaS platforms (for example Salesforce, Microsoft Copilot), cloud-built agents (AWS, Azure, Google Cloud, Databricks), agents on employee endpoints (for example Claude Desktop, Claude Code, Cursor, GitHub Copilot), MCP servers, skills and tools, plus models and AI supply chain artifacts. Consumer AI web apps in the browser are not documented.
No. Noma says policies are enforced through supported gateways, agent hooks, EDR, MDM and direct APIs, and that an MCP gateway is not required.
Noma does not publish pricing on its site. AWS Marketplace shows a $200,000 12-month platform contract and a usage-based listing at $8 to $16 per million tokens scanned, $130 per asset and $650 per red teaming test.
Yes. Agent Access Control sets each agent or MCP connection to Approved, Requires Review or Blocked, and can approve or block individual tools on a server.
No acquisition was found. Its most recent confirmed round is a $100M Series B announced July 31, 2025.