Opal Security is an access governance platform that controls who and what can reach an organisation’s systems: employees and contractors, non-human identities (NHIs) such as service accounts and API tokens, and AI agents. It sits on top of the identity provider a company already runs, such as Okta or Entra ID, and tackles the backlog of access requests, standing privilege and slow access reviews.
The platform combines an access graph, policy as code, just-in-time access and access reviews, with an AI agent called Paladin making routine access decisions and explaining each one. A separate product, Opal Zero, applies the same approach to AI agents.
At A Glance
| Vendor | Opal Security |
|---|---|
| Category | Access governance, just-in-time access, AI agent access control |
| Identities covered | Employees, contractors, external partners, service accounts, CI/CD pipelines, API tokens, AI agents |
| Deployment | Opal Cloud, or self-hosted on a customer VM or Kubernetes |
| Agentic AI | Paladin, Agents in Opal, Opal Zero, Gateway Sync, Opal MCP |
| Certifications | SOC 2 Type 2, with an independent penetration test every year |
| Funding | A reported $59 million raised in total |
| Customers | Databricks, Cloudflare, Notion, Figma, Elastic |
What Opal Does
Access requests now arrive faster than people can answer them, and AI agents make the volume worse. The usual responses, granting broad standing access or making everyone wait for a ticket, both leave risk or friction behind.
Opal organises its answer as See, Encode, Enforce. See builds one access graph of every identity, entitlement and access path, so teams know what exists. Encode turns access rules into versioned, reusable policy. Enforce hands decisions to Paladin, which approves safe requests, sends the rest to a person with its reasoning and removes access that is no longer needed. Grants are time-bound by default and every one leaves audit evidence. The organisation decides how much Paladin may decide alone.
Products
Access Graph And Opal Intelligence
Few security and IT teams can say who has access to what, least of all for the service accounts, pipelines and API tokens that nobody owns. Opal builds a central graph of access across identity providers, cloud platforms, infrastructure, SaaS applications and on-premises systems, covering human and non-human identities in one view. Opal Intelligence works on top of the graph, surfacing prioritised findings such as unused entitlements, stale assignments and excessive access.
OpalQuery
Administrators and auditors use OpalQuery, an interactive explorer, to answer access questions without exporting data from many systems. Node queries list users, resources and groups; access queries return one row per principal, asset and access level, showing how each grant came about. Results show the current state of access rather than history. Questions can be typed in natural language, a feature that can be switched off, and Access Path Queries trace a person’s or agent’s access back to its source.
OpalScript And Policy Insights
When approval rules live in tickets and people’s heads, they are applied unevenly. OpalScript holds approval workflows as version-controlled code, written by hand or generated from a plain-language description. Policy Insights scores the policies themselves on hygiene, approval efficiency and provisioning speed, so teams can see which rules slow things down.
Paladin And Just-In-Time Access
Where access requests queue for hours, standing privilege piles up because removing it is manual. Paladin reviews each request, applies policy, weighs risk and records why it decided as it did. Users ask for access from Slack or the Opal interface, receive it for a limited time and lose it automatically when the work is done. Paladin is generally available and also runs in on-premises, self-hosted deployments.
Access Reviews
Many compliance teams still run certifications from spreadsheets. Access Campaigns automate, delegate and track user access reviews from the central access data. AI-Guided Access Reviews, launched in June 2026, let Paladin read each entitlement, weigh the evidence and recommend a decision, while a person signs off every certification. The result is a reported saving of up to 90 percent of review effort.
Agents In Opal
Identity teams can hand repetitive governance work to Agents in Opal, a set of Paladin-powered agents announced at Identiverse 2026 and available to customers from July 2026. They cover four areas:
| Reviews | Less effort and better quality in access certifications |
|---|---|
| Requests | Faster fulfilment with policy checked on every grant |
| Intelligence | Analysis of access data to find risk |
| Policy | Help creating and refining access rules |
How It Works
| Component | What it does |
|---|---|
| Identity layer | Opal connects to the existing identity provider and syncs the organisation’s identities, then provisions and removes access directly in connected systems, such as IdP groups, cloud roles and application permissions. Traffic does not route through Opal. |
| Deployment | Customers can use Opal Cloud or self-host under an Opal licence on their own virtual machine, on-premises or in an existing Kubernetes cluster, with an AWS Application Load Balancer as a documented option. |
| Cloud hosting | Opal Cloud runs on AWS. Traffic over public networks uses TLS 1.2 or later, or VPN connections, and stored data is encrypted with AWS KMS. |
| Connectors | More than 250 built-in integrations, plus a guide for building custom connectors. |
| Automation | Access rules enforce policy at scale, a self-service catalogue lists what can be requested, and Terraform, a REST API and the Opal CLI let teams manage Opal as code. |
| Audit | Each Paladin decision keeps a rationale that can be retrieved later. |
Identities And Systems Covered
| People | Employees, contractors and external partners |
|---|---|
| Non-human identities | Service accounts, CI/CD pipelines and API tokens |
| AI agents | Agents, sub-agents and coding agents, plus copilots and autonomous systems, each treated as its own identity type |
| Cloud and data platforms | AWS, Google Cloud Platform, Snowflake, Databricks and Clickhouse |
| Developer and infrastructure tools | GitHub, GitLab, Vercel, Tailscale, Teleport, Twingate, Datadog and Grafana |
| Business applications | Salesforce, NetSuite, Oracle Fusion Cloud, Coupa, Docusign, Box, Hubspot, Zendesk, Zoom and Tableau |
| AI platforms | Anthropic Platform (Claude deployments, API keys and managed agents), OpenAI Platform, Cursor, Devin AI and AWS Bedrock AgentCore |
Agentic AI Capabilities
Opal Zero
Opal Zero is a separate product for organisations deploying AI and coding agents. Launched on 17 September 2026 and generally available from the end of that month, it defines least privilege for agents as three conditions together: grants that expire, an owner for every agent, and a decision taken at the moment of access instead of at provisioning time. Its inventory pulls in agents from Okta, Entra, Anthropic, AWS Bedrock AgentCore, OpenAI and Cursor, and ownership and purpose are required when an agent is ingested.
Paladin Decisioning For Agents
For agents, Paladin adds context to the security team’s written policy: who owns the agent, what it is for and what it has touched before. A request that fits that history is treated differently from one that does not.
Risk Center
The Risk Center maps the agent estate before policy is applied. It flags unowned and dormant agents, standing privileges, over-scoped access and agents without a documented purpose, and routes each finding to the owner with the fix attached.
Gateway Sync
Opal Zero adds no proxy of its own and issues no long-lived credentials. Gateway Sync writes each decision into a gateway the organisation already runs as scoped, time-bound policy, and that gateway brokers the agent’s access. Databricks Unity Gateway and AWS Bedrock AgentCore Gateway are supported at launch, with any MCP gateway in scope.
Model-Provider Controls And Opal MCP
Opal Zero brings provider-side controls into the same graph, including the Claude Compliance API, admin controls for Claude Code, Enterprise-Managed Authorization for MCP connectors and Okta Cross App Access. Opal also publishes its own MCP interface, which agents reach through Okta Cross App Access or OAuth instead of a standing key.
Integrations
| Area | Integrations |
|---|---|
| Identity providers | Okta, Okta CIAM, Microsoft Entra ID (Azure), Google, Active Directory, LDAP, Duo |
| HR systems | Workday, Rippling and other HRIS |
| Ticketing | Freshservice, Jira, Linear, Notion, ServiceNow, Shortcut |
| Messaging and on-call | Slack, Microsoft Teams, Google Chat, Lark, PagerDuty, Opsgenie |
Company Background
| Headquarters | San Francisco, United States |
|---|---|
| Leadership | Howard Ting (CEO since December 2025), Alex Pien (CTO), Sameer Mehta (CPO) |
| Funding | Investors include Greylock Partners, Battery Ventures, Box Group, SVCI and Cambium Capital; a $22 million round was led by Battery Ventures in December 2023 |
| Recognition | Named to Notable Capital’s Rising in Cyber 2026 list |
Note: This overview is compiled from Opal Security’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.