Opal Security is an AI-native access governance platform that controls who and what can access an organisation’s systems, covering employees, service accounts and other non-human identities (NHIs), and AI agents. Headquartered in San Francisco and led by CEO Howard Ting, who joined in December 2025, Opal combines an access graph, policy-as-code, just-in-time access and access reviews with an AI agent called Paladin that evaluates access decisions. In September 2026 the company launched Opal Zero, a separate product designed to bring least privilege to AI agents. This overview explains how the platform works and what it covers, based on Opal’s public product information and announcements.
At A Glance
| Vendor | Opal Security, San Francisco |
|---|---|
| Category | Identity and access governance, just-in-time access, AI agent access control |
| Identities covered | Employees, service accounts and other NHIs, AI agents, sub-agents and coding agents |
| Deployment | SaaS |
| AI engine | Paladin, an AI agent that reasons over access requests and reviews |
| Agentic AI | Opal Zero (generally available end of September 2026), Agents in Opal, Gateway Sync |
| Funding | $23 million raised in June 2026; investors include Greylock Partners and Battery Ventures |
What The Platform Does
Opal’s workforce platform brings access visibility, policy, requests and reviews together on one access graph, with AI applied across each stage.
Access Graph And Visibility
Opal builds a central access graph across identity providers, cloud platforms, infrastructure, SaaS applications and on-premises systems. Security teams can see every identity, human or non-human, and what it can reach, from a single view. Opal Intelligence surfaces prioritised insights such as unused entitlements, stale assignments and excessive access.
Policy-As-Code
Access rules are expressed as policy-as-code, so that who can request what, who must approve it and for how long it lasts are encoded, versioned and applied consistently. Policy Insights scores an organisation’s policies on hygiene, approval efficiency and how quickly access is provisioned, giving a feedback loop on the rules themselves.
Just-In-Time Access
Users request access when they need it, and Opal grants it for a limited period before revoking it automatically. Requests can be made from collaboration tools such as Slack as well as the Opal interface. Paladin evaluates requests against policy and context, approves what is clearly safe, and escalates only the requests that need a human decision. Opal reports that Databricks runs 86,000 just-in-time access requests through the platform.
Access Reviews
Opal automates user access reviews from the centralised access data. AI-Guided Access Reviews, launched in June 2026 as part of Agents in Opal, use Paladin to reason over each item, make recommendations and handle routine decisions, while a human signs off every certification. Opal states that this can reduce review effort by up to 90 percent.
Agents In Opal
Agents in Opal is a set of AI agents built into the platform and powered by Paladin. Announced at Identiverse 2026 and available to customers from July 2026, they work across four areas:
| Access reviews | Reducing review effort and improving review quality |
|---|---|
| Access requests | Speeding up fulfilment while enforcing policy on every grant |
| Intelligence | Analysing access data to identify risk |
| Policy | Helping teams create and refine access policy |
Every Paladin decision carries a documented rationale, so the reason an identity held a permission can be retrieved later for audit.
Opal Zero
Opal Zero is a separate product focused specifically on AI agent identities. It defines least privilege for agents as three conditions that must hold together: grants that expire, an owner for every agent, and a decision made at the moment of access rather than at provisioning time.
Agent Inventory
Opal Zero ingests agents from Okta, Microsoft Entra ID, Anthropic, AWS Bedrock, OpenAI and Cursor, and maps each one to an owner and a documented purpose. Agents, sub-agents and coding agents are treated as distinct identity types. Ownership and purpose are required at ingestion.
Paladin Decisioning
Paladin evaluates each agent request against the security team’s written policy and the organisation’s context, including who owns the agent, what it is for and what it has accessed before. It approves clearly safe requests, escalates those that need a human, and revokes access as soon as it is no longer needed.
Risk Center
The Risk Center describes the agent estate before policy is applied, flagging unowned agents, dormant identities, standing privileges, over-scoped access and agents with no documented purpose.
Gateway Sync
Opal Zero does not add its own proxy to the request path or issue its own long-lived credentials. Instead, Gateway Sync writes scoped, expiring permissions into a gateway the organisation already operates, which then brokers the agent’s access. Databricks Unity Gateway and AWS Bedrock AgentCore Gateway are supported at launch, and Opal states that any MCP gateway is in scope.
Model-Provider Integrations
Opal Zero connects to provider-side controls so they appear in the same access graph, including the Claude Compliance API, admin controls for Claude Code, Enterprise-Managed Authorization for MCP connectors, and Okta Cross App Access.
Availability And Pricing
Opal Zero reached general availability at the end of September 2026. Opal published a launch price of $30,000 for 12 months, available until 30 December 2026. Design partners included Instacart, Faire, Databricks and Superhuman.
How The Architecture Works
| Component | What it does |
|---|---|
| Central control plane | Opal runs as SaaS and holds the access graph, policies, request workflows, review campaigns and audit history. |
| Native enforcement | Opal provisions and removes access directly in connected systems, such as identity provider groups, cloud roles and application permissions, rather than routing traffic through Opal. |
| Gateway-based agent enforcement | For AI agents, Opal Zero enforces decisions through existing agent and MCP gateways via Gateway Sync, keeping Opal out of the data path. |
| Human in the loop | Paladin automates routine decisions, but escalations and access review certifications are signed off by people. |
Integrations
| Area | Documented integrations |
|---|---|
| Identity providers | Okta, Microsoft Entra ID |
| AI platforms and agents | Anthropic (including Claude Code), OpenAI, AWS Bedrock, Cursor |
| Agent and MCP gateways | Databricks Unity Gateway, AWS Bedrock AgentCore Gateway, other MCP gateways |
| Collaboration | Slack for access requests and approvals |
| Systems governed | Cloud platforms, infrastructure, SaaS applications and on-premises systems |
Use Cases And Industries
Opal publishes use cases for just-in-time access to replace standing privileges, automating user access reviews, governing service accounts and other NHIs, enforcing least privilege for AI agents and coding agents, and policy-as-code access governance. Its customers are concentrated among technology and AI-focused companies, including Databricks, Notion, Cloudflare, Scale AI, CoreWeave, Figma, Elastic, Grammarly, Mercari, Superhuman and Verily. Published results include an 88 percent reduction in standing access at Chronosphere.