P0 Security is a runtime access platform that controls what human users, machines and AI agents can access and do across cloud and on-premises production infrastructure. Based in San Francisco and led by co-founder and CEO Shashwat Sehgal, the company describes its product as an Authz Control Plane: a central authorization layer that replaces standing permissions and static credentials with just-in-time, just-enough privilege that expires automatically. P0 works with an organisation’s existing identity provider and native cloud APIs rather than routing traffic through a vault or proxy, and records every grant for audit. This overview explains how the platform works and what it covers, based on P0’s public product documentation and announcements.
At A Glance
| Vendor | P0 Security, San Francisco |
|---|---|
| Category | Just-in-time privileged access, NHI lifecycle management, runtime authorization for AI agents |
| Identities covered | Human users, service accounts, workloads and first-party AI agents |
| Deployment | SaaS control plane using native cloud APIs; an optional security perimeter component runs in the customer’s own cloud |
| Infrastructure covered | AWS, Google Cloud, Microsoft Azure, Kubernetes, servers over SSH and RDP in cloud or on-premises, databases and data warehouses |
| Agentic AI | Authz Control Plane for Agents (generally available February 2026); agentic runtime access control (June 2026) |
| Recognition | Named to the 2026 Cyber60 list by Lightspeed, Fortune and AWS Startups |
What The Platform Does
P0 organises the platform around three activities: discovering where privilege exists, controlling access at runtime, and proving what happened.
Inventory And Posture
P0 connects to cloud environments and builds an inventory of identities and their privileges across AWS, Google Cloud and Azure. It detects posture risks such as stale credentials, overly broad roles and embedded secrets, and maps which identities can reach critical data stores such as object storage and managed databases. P0 also provides SSH security assessments that report risk findings for servers managed through AWS Systems Manager and for self-hosted servers.
Just-In-Time Access
Just-in-time access is P0’s core workflow. Users request access from the tools they already use, including Slack, Microsoft Teams, a command-line interface and the web app. Requests are routed to approvers, who can set an expiry time, and P0 then creates a short-lived, narrowly scoped role or native entitlement directly in the target system. When the time window ends, P0 revokes the access automatically. Policies can auto-approve requests in defined conditions, for example during an on-call escalation in PagerDuty, and break-glass access is supported for emergencies. P0 reports a typical time of around five minutes from request to access.
Keyless SSH And Server Access
P0 provides just-in-time SSH and sudo access to Linux servers without shared keys or a bastion host, using the cloud provider’s own connection services, AWS Systems Manager, Google Identity-Aware Proxy and Azure Bastion, for cloud servers. The p0 ssh command lets engineers target servers by hostname, instance ID or tag. Since late 2025, P0 has extended just-in-time access to on-premises SSH and RDP targets for hybrid environments.
Access Reviews And Audit
P0 automates periodic access reviews and links every grant back to a responsible user or team. Every request, approval, grant and revocation is logged, and audit data can be sent to SIEM platforms. P0 positions these controls as support for SOC 2 and ISO 27001 audits.
NHI Lifecycle Management
P0 applies the same just-in-time model to non-human identities. NHI Lifecycle Management, generally available since February 2026, covers:
| Discovery | Continuous discovery of service accounts and workload identities across cloud environments, with mapping of what each can access |
|---|---|
| Ownership | Assigning each NHI to an accountable owner or team |
| Intent | Defining the intended usage of each NHI when it is provisioned |
| Credential rotation | Rotating NHI credentials automatically |
| Programmatic just-in-time access | Service accounts and workloads request access through platform integrations and receive time-bound, purpose-specific entitlements instead of standing permissions |
| Audit | Session-level audit trails for machine access |
How The Architecture Works
| Component | What it does |
|---|---|
| Control plane | The P0 platform runs as SaaS. It stores policy, handles requests and approvals, and calls the native IAM APIs of each connected system to create and remove entitlements. |
| Native enforcement | Access is enforced by the target systems themselves, using IAM roles, database grants, Kubernetes role bindings and similar native controls. P0 does not sit in the data path for most access, and users keep their existing identities from the corporate directory. |
| Security perimeter | For connected cloud environments and identity providers, P0 documents a security perimeter component that runs in the customer’s own cloud, for example as a Cloud Run service in Google Cloud, and handles provisioning operations within the customer’s environment. |
| Identity provider as source of truth | P0 ties all access back to the corporate directory, using Okta, Microsoft Entra ID, Google Workspace or LDAP, so that group membership and user status drive approvals and reviewer assignments. |
Agentic AI Capabilities
P0 extends its authorization model to AI agents, keeping track of both the agent and the human or system that started it.
Authz Control Plane For Agents
Generally available since February 2026, the Authz Control Plane for Agents replaces persistent agent permissions with time-bound, purpose-specific authorization. First-party agents are authorized at runtime during tool execution, and every action is attributed to the originating user or system. P0 introduced early-access controls for agents built on AWS Bedrock and Google Vertex AI in late 2025.
Runtime Access Control Across The Action Chain
In June 2026, P0 announced agentic runtime access control capabilities that evaluate authorization when access is requested and enforce policy across the full action chain, from the originator through the agent, tools and gateways to the target system. Agents keep their existing identities, while P0 preserves originator and agent context so each decision can be explained afterwards. P0 has documented deploying this layer across three AI agent architectures in a customer’s Google Cloud environment, using the customer’s identity provider for inbound authentication and native Google Cloud mechanisms for outbound access.
Coding Agents
P0 addresses the use of coding agents such as Claude Code by engineers, where an agent may run scripts on remote servers or query production databases. Access requested through an agent is subject to the same just-in-time and least-privilege controls as direct human access.
Integrations
| Area | Documented integrations |
|---|---|
| Cloud platforms | AWS, Google Cloud, Microsoft Azure |
| Servers | Linux servers over SSH and sudo, Windows over RDP, in cloud or on-premises |
| Databases and data | PostgreSQL, Cloud SQL, Snowflake and other databases |
| Containers and code | Kubernetes and source code repositories |
| Directories | Okta, Microsoft Entra ID, Google Workspace, LDAP and Active Directory |
| Workflow and incident tools | Slack, Microsoft Teams, PagerDuty, Jira |
| Agentic platforms | AWS Bedrock, Google Vertex AI |
| Other | SIEM export, CI/CD, and an SDK for adding just-in-time access to any application |
Use Cases And Industries
P0 publishes use cases for just-in-time access to production cloud infrastructure, keyless SSH, on-call and break-glass access, temporary access to customer data in cloud environments and data warehouses, automated access reviews for compliance, NHI governance, and runtime authorization for AI agents. It reports deployments in financial services, healthcare, SaaS and cloud-native companies. Customers referenced in P0 announcements include Splunk and Resolve AI.