Non-human identity security tools compared
SHORT ANSWER
Editorial assessment · Desk research from public vendor material · Reviewed 2026-09-29
Ranking
| Rank | Product | Total | Best for |
|---|---|---|---|
| 1 | Oasis Security | 7.5 / 10 | Broad inventory and ownership across clouds, SaaS, vaults and AI services |
| 2 | Entro Security | 7.3 / 10 | Finding secrets in code, pipelines and chat tools and mapping them to owners |
| 3= | Aembit | 7.0 / 10 | Replacing stored secrets with credentials issued per request |
| 3= | Silverfort | 7.0 / 10 | Fencing Active Directory service accounts that cannot be changed |
| 5 | NewCore | 6.7 / 10 | Organizations whose non-human identity problem is mostly AI agents |
| 6 | CyberArk Secrets Manager | 5.8 / 10 | Central rotation of application secrets, including self-hosted and open-source options |
How the scores are weighted
- c1 Discovery breadth 25%
- c2 Owner for every identity 15%
- c3 Removing long-lived secrets 25%
- c4 Runtime enforcement 15%
- c5 AI agents as non-human identities 10%
- c6 Maturity and buying clarity 10%
Totals are the weighted average of the criterion scores, computed from the weights shown. Nothing is adjusted by hand.
Scorecard
| Criterion | Oasis Security7.5 / 10 | Entro Security7.3 / 10 | Aembit7.0 / 10 | Silverfort7.0 / 10 | NewCore6.7 / 10 | CyberArk Secrets Manager5.8 / 10 |
|---|---|---|---|---|---|---|
| c1Discovery breadth25%Clouds, SaaS, code, pipelines and vaults searched for service accounts, keys and tokens. | 9Highest in setIntegrations span Azure, AWS, GCP, Ping, Okta, Active Directory, HashiCorp, cloud key vaults, Snowflake, Databricks, GitHub, Salesforce and AI services. Source: Oasis Security | 9Highest in setSearches clouds, code, CI/CD, on-prem and collaboration tools, including Vault, AWS Secrets Manager, Azure Key Vault, GitHub, GitLab, Kubernetes, Jenkins, Slack and ServiceNow. Source: Entro Security | 4Gap in public materialLowest in setBuilt to issue access, not to search for existing keys; discovery of existing secrets is not described. Source: Aembit | 7Gap in public materialDiscovers AD service accounts and cloud NHIs; code repositories and vaults are not listed. | 5Gap in public materialDiscovery covers people, agents, systems and apps across Okta, AWS, Microsoft, BambooHR, Salesforce and Google. Code repositories and vaults are not listed. Source: NewCore: Identity Discovery | 5Manages the secrets it holds and syncs with AWS Secrets Manager and Azure Key Vault through Secrets Hub; broad discovery is not the focus. Source: CyberArk secrets management |
| c2Owner for every identity15%Can each non-human identity be tied to a responsible person? | 8Assigns identities to owners with context on what each does. Source: Oasis Security | 9Highest in setMaps every NHI and secret to a human owner. Source: Entro Security | 6Gap in public materialBlended Identity ties agent access to the human operating it; ownership of other workloads is not described. Source: Aembit: IAM for agentic AI | 8Maps human owners, sources, destinations and privilege levels for each account. | 8Ties every agent action to an accountable human or team. Source: NewCore: Agent Guardian | 4Gap in public materialLowest in setOwnership mapping is not described on the page reviewed. Source: CyberArk secrets management |
| c3Removing long-lived secrets25%Does the product replace or rotate long-lived secrets, or only report them? | 7Safe secret rotation and time-bound access for agents. Source: Oasis Security | 6Rotation, vaulting and just-in-time access are listed; the emphasis is on finding and tracking secrets. Source: Entro Security | 9Highest in setSecretless access: credentials are delivered just in time, per task. Source: Aembit | 5Lowest in setConstrains existing accounts rather than replacing their secrets. | 7Replaces standing credentials with short-lived, scoped ones for agents; the same claim is not made for existing service accounts. Source: NewCore: Agent Guardian | 8Centrally rotates and manages credentials for applications, DevOps tools and pipelines. Source: CyberArk secrets management |
| c4Runtime enforcement15%Can abnormal use be blocked as it happens, not only reported? | 6Gap in public materialThreat and anomaly detection; inline blocking is not described. Source: Oasis Security | 6Gap in public materialDetects abuse in real time through its NHIDR engine; inline blocking is not described. Source: Entro Security · Entro: AI agents | 8Conditional access policies evaluated at request time, including security posture. Source: Aembit | 9Highest in setBehavioral baseline per service account and "virtual fences" enforced inline. | 8Per-call policy for agents, with the option to end a single path. Source: NewCore: Agent Guardian | 5Gap in public materialLowest in setControls how applications fetch secrets; behavioral blocking is not described. Source: CyberArk secrets management |
| c5AI agents as non-human identities10%Are AI agents covered alongside service accounts and keys? | 8AI security posture and Agentic Access Management for agents. Source: Oasis Security | 8Inventories agents and links them to the secrets and identities they use. Source: Entro: AI agents | 8Covers AI agents and MCP servers through its MCP Identity Gateway. Source: Aembit: IAM for agentic AI | 8Covers AI agents alongside service accounts and tokens. Source: Silverfort: non-human identity security · Silverfort: AI agent security | 9Highest in setTreats on-behalf-of, delegated and autonomous agents as first-class identities. Source: NewCore home page | 3Gap in public materialLowest in setAI agents are not mentioned on the page reviewed. Source: CyberArk secrets management |
| c6Maturity and buying clarity10%Time in market, certifications, free options and ownership status. | 6States SOC 2 and ISO 27001; now part of Cyera, which buyers should factor into roadmap questions. Source: Oasis Security | 5Gap in public materialPricing and certifications are not on the pages reviewed. Source: Entro Security | 8Free-forever tier; states SOC 2 Type II and ISO 27001:2022. Source: Aembit · Aembit: IAM for agentic AI | 6Gap in public materialEstablished vendor; pricing not published. Source: Silverfort platform | 4Lowest in setLaunched June 2026; no pricing or named customers. | 9Highest in setSaaS, self-hosted and open-source (Conjur) options; named overall leader in KuppingerCole's 2025 Leadership Compass for enterprise secrets management. Source: CyberArk secrets management |
Oasis Security
7.5 / 10
c1 · 25%
Discovery breadth
9Highest in set
Integrations span Azure, AWS, GCP, Ping, Okta, Active Directory, HashiCorp, cloud key vaults, Snowflake, Databricks, GitHub, Salesforce and AI services.
Source: Oasis Security
c2 · 15%
Owner for every identity
8
Assigns identities to owners with context on what each does.
Source: Oasis Security
c3 · 25%
Removing long-lived secrets
7
Safe secret rotation and time-bound access for agents.
Source: Oasis Security
c4 · 15%
Runtime enforcement
6Gap in public material
Threat and anomaly detection; inline blocking is not described.
Source: Oasis Security
c5 · 10%
AI agents as non-human identities
8
AI security posture and Agentic Access Management for agents.
Source: Oasis Security
c6 · 10%
Maturity and buying clarity
6
States SOC 2 and ISO 27001; now part of Cyera, which buyers should factor into roadmap questions.
Source: Oasis Security
Entro Security
7.3 / 10
c1 · 25%
Discovery breadth
9Highest in set
Searches clouds, code, CI/CD, on-prem and collaboration tools, including Vault, AWS Secrets Manager, Azure Key Vault, GitHub, GitLab, Kubernetes, Jenkins, Slack and ServiceNow.
Source: Entro Security
c2 · 15%
Owner for every identity
9Highest in set
Maps every NHI and secret to a human owner.
Source: Entro Security
c3 · 25%
Removing long-lived secrets
6
Rotation, vaulting and just-in-time access are listed; the emphasis is on finding and tracking secrets.
Source: Entro Security
c4 · 15%
Runtime enforcement
6Gap in public material
Detects abuse in real time through its NHIDR engine; inline blocking is not described.
Source: Entro Security · Entro: AI agents
c5 · 10%
AI agents as non-human identities
8
Inventories agents and links them to the secrets and identities they use.
Source: Entro: AI agents
c6 · 10%
Maturity and buying clarity
5Gap in public material
Pricing and certifications are not on the pages reviewed.
Source: Entro Security
Aembit
7.0 / 10
c1 · 25%
Discovery breadth
4Gap in public materialLowest in set
Built to issue access, not to search for existing keys; discovery of existing secrets is not described.
Source: Aembit
c2 · 15%
Owner for every identity
6Gap in public material
Blended Identity ties agent access to the human operating it; ownership of other workloads is not described.
Source: Aembit: IAM for agentic AI
c3 · 25%
Removing long-lived secrets
9Highest in set
Secretless access: credentials are delivered just in time, per task.
Source: Aembit
c4 · 15%
Runtime enforcement
8
Conditional access policies evaluated at request time, including security posture.
Source: Aembit
c5 · 10%
AI agents as non-human identities
8
Covers AI agents and MCP servers through its MCP Identity Gateway.
Source: Aembit: IAM for agentic AI
c6 · 10%
Maturity and buying clarity
8
Free-forever tier; states SOC 2 Type II and ISO 27001:2022.
Source: Aembit · Aembit: IAM for agentic AI
Silverfort
7.0 / 10
c1 · 25%
Discovery breadth
7Gap in public material
Discovers AD service accounts and cloud NHIs; code repositories and vaults are not listed.
c2 · 15%
Owner for every identity
8
Maps human owners, sources, destinations and privilege levels for each account.
c3 · 25%
Removing long-lived secrets
5Lowest in set
Constrains existing accounts rather than replacing their secrets.
c4 · 15%
Runtime enforcement
9Highest in set
Behavioral baseline per service account and "virtual fences" enforced inline.
c5 · 10%
AI agents as non-human identities
8
Covers AI agents alongside service accounts and tokens.
Source: Silverfort: non-human identity security · Silverfort: AI agent security
c6 · 10%
Maturity and buying clarity
6Gap in public material
Established vendor; pricing not published.
Source: Silverfort platform
NewCore
6.7 / 10
c1 · 25%
Discovery breadth
5Gap in public material
Discovery covers people, agents, systems and apps across Okta, AWS, Microsoft, BambooHR, Salesforce and Google. Code repositories and vaults are not listed.
Source: NewCore: Identity Discovery
c2 · 15%
Owner for every identity
8
Ties every agent action to an accountable human or team.
Source: NewCore: Agent Guardian
c3 · 25%
Removing long-lived secrets
7
Replaces standing credentials with short-lived, scoped ones for agents; the same claim is not made for existing service accounts.
Source: NewCore: Agent Guardian
c4 · 15%
Runtime enforcement
8
Per-call policy for agents, with the option to end a single path.
Source: NewCore: Agent Guardian
c5 · 10%
AI agents as non-human identities
9Highest in set
Treats on-behalf-of, delegated and autonomous agents as first-class identities.
Source: NewCore home page
c6 · 10%
Maturity and buying clarity
4Lowest in set
Launched June 2026; no pricing or named customers.
CyberArk Secrets Manager
5.8 / 10
c1 · 25%
Discovery breadth
5
Manages the secrets it holds and syncs with AWS Secrets Manager and Azure Key Vault through Secrets Hub; broad discovery is not the focus.
Source: CyberArk secrets management
c2 · 15%
Owner for every identity
4Gap in public materialLowest in set
Ownership mapping is not described on the page reviewed.
Source: CyberArk secrets management
c3 · 25%
Removing long-lived secrets
8
Centrally rotates and manages credentials for applications, DevOps tools and pipelines.
Source: CyberArk secrets management
c4 · 15%
Runtime enforcement
5Gap in public materialLowest in set
Controls how applications fetch secrets; behavioral blocking is not described.
Source: CyberArk secrets management
c5 · 10%
AI agents as non-human identities
3Gap in public materialLowest in set
AI agents are not mentioned on the page reviewed.
Source: CyberArk secrets management
c6 · 10%
Maturity and buying clarity
9Highest in set
SaaS, self-hosted and open-source (Conjur) options; named overall leader in KuppingerCole's 2025 Leadership Compass for enterprise secrets management.
Source: CyberArk secrets management
What are non-human identities?
Non-human identities, or NHIs, are the accounts and credentials that software uses to reach other software: service accounts, API keys, OAuth tokens, certificates, secrets in pipelines and, increasingly, AI agents. OWASP puts it simply: production environments contain a large number of applications "which need to be identified".
Are machine identities the same as non-human identities?
The terms overlap. Machine identity is the older term and is often used for certificates and keys that identify servers and workloads. Non-human identity is the broader term used by most of the vendors on this page, covering service accounts, tokens and agents as well. This page treats machine identities as part of the non-human set.
What are the main risks? The OWASP NHI Top 10
OWASP's Non-Human Identities Top 10 for 2025 is the common reference list:
- Improper Offboarding
- Secret Leakage
- Vulnerable Third-Party NHI
- Insecure Authentication
- Overprivileged NHI
- Insecure Cloud Deployment Configurations
- Long-Lived Secrets
- Environment Isolation
- NHI Reuse
- Human Use of NHI
The scorecard's discovery and ownership criteria address offboarding, leakage and reuse; removing long-lived secrets addresses items 2 and 7; runtime enforcement addresses over-privilege and misuse.
Finding secrets or replacing them: two kinds of NHI tool
The most useful distinction in this category is between tools that find and manage the non-human identities you already have, and tools that issue new, short-lived credentials so that fewer long-lived ones exist. Oasis, Entro and Silverfort are mainly in the first group. Aembit and NewCore are mainly in the second, and CyberArk sits in between by holding and rotating secrets centrally. Most organizations need both: an inventory of what exists today and a way to stop creating more.
What does machine identity governance involve?
In practice: an owner for every non-human identity, a reason it exists, a record of what it can reach, rotation or expiry for its credentials, and removal when the owner or the purpose goes away. The first item on the OWASP list, improper offboarding, is what happens when that last step is missed.
Managing and securing non-human identities: where to start
- Inventory service accounts, keys and tokens across clouds, code and SaaS.
- Assign an owner to each; flag the ones nobody claims.
- Remove or fence the unused and the over-privileged.
- Move the remaining secrets to short-lived credentials where the target system allows it.
- Watch the rest for abnormal use, and block it where you can.
- Treat new AI agents as non-human identities from the day they are created.
A note on vendors that have changed hands
Astrix Security, a well-known NHI vendor, is now part of Cisco and ended standalone sales of new licenses on 30 June 2026, so it is not scored here. Oasis Security is now part of Cyera; it is scored, and buyers should ask how the product will be sold and supported.
The six tools, one by one
Oasis Security
Non-human identity security, now part of Cyera
Best for: Broad inventory and ownership across clouds, SaaS, vaults and AI services
LEADS ON
c1 Discovery breadth 9
Integrations span Azure, AWS, GCP, Ping, Okta, Active Directory, HashiCorp, cloud key vaults, Snowflake, Databricks, GitHub, Salesforce and AI services.
Source: Oasis Security
c2 Owner for every identity 8
Assigns identities to owners with context on what each does.
Source: Oasis Security
TRAILS ON
c4 Runtime enforcement 6
Threat and anomaly detection; inline blocking is not described.
Source: Oasis Security
c6 Maturity and buying clarity 6
States SOC 2 and ISO 27001; now part of Cyera, which buyers should factor into roadmap questions.
Source: Oasis Security
Oasis Security alternatives · Compare Oasis Security head to head
Entro Security
Non-human identity and secrets security
Best for: Finding secrets in code, pipelines and chat tools and mapping them to owners
LEADS ON
c1 Discovery breadth 9
Searches clouds, code, CI/CD, on-prem and collaboration tools, including Vault, AWS Secrets Manager, Azure Key Vault, GitHub, GitLab, Kubernetes, Jenkins, Slack and ServiceNow.
Source: Entro Security
c2 Owner for every identity 9
Maps every NHI and secret to a human owner.
Source: Entro Security
TRAILS ON
c6 Maturity and buying clarity 5
Pricing and certifications are not on the pages reviewed.
Source: Entro Security
c3 Removing long-lived secrets 6
Rotation, vaulting and just-in-time access are listed; the emphasis is on finding and tracking secrets.
Source: Entro Security
Entro Security alternatives · Compare Entro Security head to head
Aembit
Access management for workloads and AI agents
Best for: Replacing stored secrets with credentials issued per request
LEADS ON
TRAILS ON
c1 Discovery breadth 4
Built to issue access, not to search for existing keys; discovery of existing secrets is not described.
Source: Aembit
c2 Owner for every identity 6
Blended Identity ties agent access to the human operating it; ownership of other workloads is not described.
Source: Aembit: IAM for agentic AI
Silverfort
Inline identity protection across on-prem, cloud and agents
Best for: Fencing Active Directory service accounts that cannot be changed
LEADS ON
c4 Runtime enforcement 9
Behavioral baseline per service account and "virtual fences" enforced inline.
c2 Owner for every identity 8
Maps human owners, sources, destinations and privilege levels for each account.
TRAILS ON
c3 Removing long-lived secrets 5
Constrains existing accounts rather than replacing their secrets.
c6 Maturity and buying clarity 6
Established vendor; pricing not published.
Source: Silverfort platform
NewCore
Identity provider for people and AI agents, launched June 2026
Best for: Organizations whose non-human identity problem is mostly AI agents
LEADS ON
c5 AI agents as non-human identities 9
Treats on-behalf-of, delegated and autonomous agents as first-class identities.
Source: NewCore home page
c2 Owner for every identity 8
Ties every agent action to an accountable human or team.
Source: NewCore: Agent Guardian
TRAILS ON
c6 Maturity and buying clarity 4
Launched June 2026; no pricing or named customers.
c1 Discovery breadth 5
Discovery covers people, agents, systems and apps across Okta, AWS, Microsoft, BambooHR, Salesforce and Google. Code repositories and vaults are not listed.
Source: NewCore: Identity Discovery
CyberArk Secrets Manager
Secrets management for applications and pipelines
Best for: Central rotation of application secrets, including self-hosted and open-source options
LEADS ON
c6 Maturity and buying clarity 9
SaaS, self-hosted and open-source (Conjur) options; named overall leader in KuppingerCole's 2025 Leadership Compass for enterprise secrets management.
Source: CyberArk secrets management
c3 Removing long-lived secrets 8
Centrally rotates and manages credentials for applications, DevOps tools and pipelines.
Source: CyberArk secrets management
TRAILS ON
c5 AI agents as non-human identities 3
AI agents are not mentioned on the page reviewed.
Source: CyberArk secrets management
c2 Owner for every identity 4
Ownership mapping is not described on the page reviewed.
Source: CyberArk secrets management
Visit CyberArk Secrets Manager
CyberArk Secrets Manager alternatives · Compare CyberArk Secrets Manager head to head
Questions
What are non-human identities?
The accounts and credentials software uses to reach other software: service accounts, API keys, OAuth tokens, certificates, secrets and AI agents.
Which NHI security tool scores highest?
Oasis Security on the weights used here, with Entro Security close behind. Aembit scores highest on removing long-lived secrets and Silverfort on runtime enforcement.
Is secrets management the same as NHI security?
It is one part. A secrets manager stores and rotates the secrets it holds. NHI security tools also look for identities and secrets that were never put in a vault, and tie them to owners.
Are AI agents non-human identities?
Yes, in the OWASP sense and in how all six vendors here describe them. They differ from service accounts in that they choose their own actions, which is why per-call checks matter more for them.
Related topics
From the blog: Seven identity security myths, checked against the published facts
Sources
- Oasis Security: https://www.oasis.security/
- Entro Security: https://entro.security/
- Entro: AI agents: https://entro.security/platform-ai-agents/
- Aembit: https://aembit.io/
- Aembit: IAM for agentic AI: https://aembit.io/iam-for-agentic-ai/
- Silverfort: non-human identity security: https://www.silverfort.com/platform/non-human-identity-security/
- Silverfort: AI agent security: https://www.silverfort.com/platform/ai-agent-security
- Silverfort platform: https://www.silverfort.com/
- CyberArk secrets management: https://www.cyberark.com/products/secrets-management/
- NewCore: Identity Discovery: https://newcore.com/platform/identity-discovery
- NewCore: Agent Guardian: https://newcore.com/platform/agent-guardian
- NewCore home page: https://newcore.com/
- NewCore launch release (PR Newswire): https://www.prnewswire.com/news-releases/newcore-emerges-from-stealth-with-66m-to-rebuild-workforce-identity-for-the-agentic-era-302799643.html
- OWASP Non-Human Identities Top 10 (2025): https://owasp.org/www-project-non-human-identities-top-10/
- Astrix Security: https://astrix.security/