Ocean
Also known as: Ocean Security, Ocean Ray
Agentic email security platform whose network of AI agents, led by an autonomous investigation engine called Ray, inspects every message in real time to catch phishing and social engineering that evade legacy filters.
Ocean is an agentic email security company headquartered in New York City and Tel Aviv, founded in 2024 by Shay Shwartz and Oran Moyal, both former Israeli military intelligence operators who later led offensive social engineering operations. It emerged from stealth with $28 million in total funding, an $8 million seed led by Picture Capital followed by a Series A led by Lightspeed Venture Partners. Its thesis is that generative AI has made fluent, context rich spear phishing cheap to produce, which breaks detection built on rules and anomalies.
Ray, Ocean's central engine, coordinates a swarm of purpose built agents that investigate every inbound message the way an analyst would, following the evidence across sender identity, content, links and infrastructure. Ocean groups the work in three. Agentic protection covers phishing, BEC, vendor compromise and malware.
Agentic automation triages reported phishing, handles quarantine release requests and removes campaigns, mostly without human intervention, and some reported emails are left for manual review. On demand agentic investigation serves incident responders.
It connects to Microsoft 365 and Google Workspace by API, and an MCP server in early access lets assistants such as Claude, ChatGPT and Cursor query the threats caught, the emails reported and the decisions made.
Ocean publishes no pricing and names no attestations, access controls, hosting regions, model providers or approval step before the agents act. The product works on the email channel only.
Vendor details
Canonical URL
https://ocean.security
Category
Security / SOC agent
Subcategory
Agentic email security
Funding status
Independent, dual headquartered in New York City and Tel Aviv, founded in 2024 by Shay Shwartz (CEO) and Oran Moyal (CTO), both former Israeli intelligence operators with offensive social engineering backgrounds. Ocean emerged from stealth in May 2026 with twenty eight million dollars total, an eight million dollar seed led by Picture Capital and a Series A led by Lightspeed Venture Partners, with Cerca Partners and angels including Assaf Rappaport of Wiz and Armis founders Yevgeny Dibrov and Nadir Izrael. The company reports seven figure revenue and roughly thirty five employees.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Ocean connects to Microsoft 365 and Google Workspace through their APIs, where it quarantines mail, remediates campaigns and releases quarantined messages, with Slack and SIEM connections named. An MCP server in early access lets assistants such as Claude, ChatGPT and Cursor query the threats caught, reported emails and the platform's decisions. The connector set is short and centered on email, and Ocean names no way to add others.
In practice
An attacker sends a flawless impersonation of a trusted colleague with no obvious anomalies. Ocean's Ray engine reads the intent behind the message and quarantines it before the employee ever acts.
A security team wastes hours chasing low value alerts while real threats blend in. Ocean investigates every email autonomously and surfaces only the messages that carry genuine malicious intent.
Employees report suspicious emails and ask for quarantined mail back all day long. Ocean triages reported phishing, handles quarantine release requests and removes campaigns without human intervention, so employees get an instant response and the SOC never touches most of them.
Sources & related URLs
Agentic Index coverage score
5.5 / 14 capabilities · 39%
| Integrations & Tool Calling | Partial |
|---|---|
|
Ocean acts inside Microsoft 365 and Google Workspace by API, quarantining mail, remediating campaigns and releasing quarantined messages, with Slack and SIEM connections named. The connector set is short and fixed, centered on email, and Ocean names no way to add others. SourceOcean, ocean.security/platformread 2026-10-06 |
|
| Workflow Orchestration | Full |
|
Ray, the central engine, coordinates a swarm of purpose built agents, among them infrastructure, file, link, identity, financial, contact, quarantine and abuse mailbox agents, that understand intent, enrich context and follow the evidence on each message, then quarantine it, remediate a campaign or release it. Several agents share each investigation, and the customer does not configure the flow. SourceOcean, ocean.security/platformread 2026-10-06 |
|
| Knowledge Grounding & RAG | Partial |
|
Investigations draw on the full context of the customer's environment and sender history, including which domains are writing for the first time. Ocean names no maintained index or graph over the customer's knowledge. SourceOcean, ocean.security/platform and /resources/blog/ocean-mcp; ocean.securityread 2026-10-06 |
|
| Human Oversight & Guardrails | Partial |
|
Some reported emails are left for manual review, while most triage, release and remediation runs without human intervention. Ocean says employees get an instant response to reported phishing and release requests and the SOC never touches them. Ocean names no approval step or admin policy that gates an action. SourceOcean, ocean.security/platform and /resources/blog/ocean-mcp; ocean.security/resources/blog/ocean-mcpread 2026-10-06 |
|
| Security, Identity & Governance | Not documented |
|
No attestation, SSO option or role model is named, and no security report is published openly. The trust center at trust.ocean.security is a separate site. SourceOcean, ocean.security, /platform and /resources/blog/ocean-mcp; trust.ocean.securityread 2026-10-06 |
|
| Observability & Auditability | Full |
|
Every verdict comes with the full reasoning behind it, every signal checked and every step taken, backed by evidence, and the threats Ocean caught, the remediation it performed on each one and the decisions it made on the customer's behalf can be queried through the MCP server. Ocean names no separate audit log export. SourceOcean, ocean.security/platform and /resources/blog/ocean-mcpread 2026-10-06 |
|
| Memory & State Persistence | Not documented |
|
By Ocean's account, the platform builds a living memory of how the organization operates and communicates, adapting from day one and constantly learning. Ocean names no memory store, scope or lifetime behind that claim. SourceOcean, ocean.security/platformread 2026-10-06 |
|
| Deployment & Data Residency | Not documented |
|
Connections to Microsoft 365 and Google Workspace run by API. No hosting region or customer environment option is named, and Ocean keeps a separate trust center at trust.ocean.security. SourceOcean, ocean.security and /platform; trust.ocean.securityread 2026-10-06 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Ray and a swarm of purpose built agents come ready to run, grouped as protection, automation and investigation, and the agents in the swarm are named, from the link, file and identity agents to the financial, contact, quarantine and abuse mailbox agents. They work as parts of one investigation under Ray, not as agents offered separately, and Ocean names no templates. SourceOcean, ocean.security/platformread 2026-10-06 |
|
| Triggers & Channel Coverage | Full |
|
Every inbound email and every phishing report or quarantine release request starts an investigation with no one at the customer asking, and each email is investigated before a compromise rather than after. The agents start on their own when messages arrive from outside, all through the one email channel. Through the MCP server, an assistant can also ask Ocean its questions on a schedule. SourceOcean, ocean.security/platform and /resources/blog/ocean-mcpread 2026-10-06 |
|
| Model Flexibility & Routing | Not documented |
|
The agents run on SLMs and LLMs that Ocean says are built specifically for email security. The models behind Ray are not named, and the customer has no model choice. SourceOcean, ocean.security/platformread 2026-10-06 |
|
| APIs, SDKs & MCP Extensibility | Partial |
|
Ocean's own MCP server lets Claude, ChatGPT, Cursor and other MCP clients query the threats caught and the remediation performed on each, reported emails, sender history, domains seen for the first time, the platform's decisions and the metrics leadership asks for, and it is listed in the usual MCP directories. It is in early access, and Ocean publishes no endpoint, auth scheme or tool list. SourceOcean, ocean.security/resources/blog/ocean-mcpread 2026-10-06 |
|
| Testing, Debugging & Optimization | Not documented |
|
For a change to the agents, Ocean names no evaluation harness, scored test cases or gate. The head to head tests customers describe, such as running Ocean against two leading email security platforms, are purchase evaluations against incumbent tools. SourceOcean, ocean.securityread 2026-10-06 |
|
| Browser & Computer Use | Not documented |
|
The agents work through the Microsoft 365 and Google Workspace APIs, and Ocean describes none driving a browser, desktop or remote computer. SourceOcean, ocean.security/platformread 2026-10-06 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Ray, Ocean's email investigation agent, can now be added from the Gemini Enterprise agent marketplace and connected to a customer's Ocean tenant. It joins over the open Agent2Agent protocol as an agent rather than a tool, so other agents hand it a task and Ray runs the full investigation, streams progress and asks a follow up question when a request is unclear. Other agents get Ray's verdict and evidence, never raw access to the email data.
Bears on: MCP / tool calling / API
View sourceOcean opened its MCP server to customers in early access, so Claude, ChatGPT, Cursor and other MCP clients can query its email security data in plain language. An agent can reach the threats Ocean caught and the remediation it performed, the emails users reported, the decisions the platform made and threat metrics over time, and combine them in one conversation with a SIEM, ticketing system or directory.
Bears on: Security / enterprise
View sourceOcean has opened direct customer access to Ray, its central autonomous AI investigation engine. While Ray previously operated strictly in the background to automatically triage and remediate emails, security teams can now query the AI directly to run on-demand investigations, reconstruct attacks, and trace threat actors.
Bears on: Security / enterprise
View sourcePricing
Not public; quoted through sales after a demo
not published
What is public
No list prices, rates or entry point are published.
Billing mechanics
Sold through sales after a demo; the billing unit is not published.
Cost watchouts
Migrating off an existing email security vendor may carry overlap or transition cost during the switch.
Variable cost rationale
No billing unit is published; as an inference, email security is commonly priced per protected mailbox, which would make cost predictable per seat.
Additional watchouts
Positioned as a replacement rather than an add on, so evaluate against the cost of the incumbent it displaces.
Sales call required
Yes, required for paid access
Free / trial
Proof of value deployments; no public free tier
Key ambiguities
No price anchor, billing unit or minimum commitment is disclosed.
Related vendors
- 7AI — Agentic SOC from the Cybereason founders: domain specialized agents…
- Abnormal AI — Behavioral AI email security with three named autonomous agents: a…
- AirMDR — AI native MDR whose AI analyst investigates every alert and writes…
- Airrived — Agentic OS that unifies SOC, GRC, IAM, vulnerability management, IT,…
- Andesite — Human-AI SOC platform where customers build their own agents and…
- Anvilogic — Agentic security operations platform that works on top of existing…
Alternatives to Ocean
The closest documented capability profiles to Ocean among security and SOC agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Equixly6.0 / 14Adds documented Security, Identity & Governance
- Terra Security6.5 / 14Adds documented Security, Identity & Governance and Testing, Debugging & Optimization
- Token Security6.5 / 14Adds documented Security, Identity & Governance
- Vulnetic5.5 / 14Adds documented Deployment & Data Residency
- Astelia8.0 / 14Adds documented Security, Identity & Governance
- Dux3.0 / 14A lighter documented profile than Ocean
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded