Wing Security finds the AI agents running inside an organisation, maps what each one can reach and cuts back the access they do not need. It is for security, exposure and AI adoption teams who have agents connecting to SaaS, code and data systems through keys and service accounts that nobody is tracking.
Its users are SOC and exposure teams, risk and AI adoption teams, and the MDR and MSSP providers that serve them. The platform grew out of an identity-centric SaaS Security Posture Management (SSPM) product, so it starts from the identities and grants behind each agent rather than from its prompts.
At A Glance
| Vendor | Wing Security (With Wings) |
|---|---|
| Category | AI agent exposure management, AI security posture management, SSPM |
| Identities covered | AI agents, service accounts, credentials, SaaS users and OAuth apps |
| Deployment | Agentless, API-first SaaS |
| Agentic AI | Agent discovery, permission governance, drift detection, Claude session monitoring |
| Customers | Carrefour, At-Bay, Monte Carlo, Apollo, Tripledot |
| Founded | 2020 |
| Market classification | Identity Security Posture Management; also NHI Platforms & Machine Identity |
What Wing Security Does
AI agents are spreading faster than anyone can approve them, and many run on long-lived keys, shared service accounts and broad delegated grants. That makes agent access, rather than SaaS misconfiguration alone, the fastest-growing exposure. Wing Security’s approach is to find every agent, including the ones nobody deployed, and map what each can reach. It follows authority through inherited and delegated grants, not just the first connection, so teams see the real reach of each agent. The platform then minimises that access and triggers approved enforcement actions. It concentrates on exposure and leaves in-flight blocking of individual actions to runtime interception tools.
Products
Observability Agent
The Observability Agent answers the first question every team asks: which agents do we have? It discovers and inventories AI agents, IDE agents and vendor-enabled agents with their identities and connections, shows each one with its owner, access and actions taken, learns how agents are used in the organisation and produces root cause graphs that show how an exposure arose.
Control Agent
The Control Agent is for teams that need agents to stay inside policy. It continuously checks roles, permissions, identity assignments, privileged access and authentication methods against identity and governance rules, flags over-privileged agents and unused rights, and catches configuration drift and unauthorised permission changes. Drift detection also compares what an agent does over time with the purpose it was set up for.
Detection, Enforcement, Mitigation And Notification Agents
These four agents handle what happens when something goes wrong.
- Detection analyses agent behaviour and identity signals for suspicious activity
- Enforcement triggers approved actions when access or behaviour leaves policy
- Mitigation reduces exposure and starts the response
- Notification routes the alert and the owner to the right team
WithWings For Claude
WithWings for Claude gives organisations adopting Anthropic’s tools visibility of Claude Managed Agents and Claude Code: who built each agent, who uses it, and what it did in each session, from instructions to actions and results. Recorded sessions are checked for unexpected actions, access to data outside the agent’s task and behaviour that does not fit its purpose.
How It Works
The platform is agentless and API-first. It connects to the systems agents reach through more than 180 native integrations and acts as one control plane between agents and those systems, working in four steps: find, map, minimise, enforce. The six security agents share one fabric.
| Component | What it does |
|---|---|
| Policy | Declarative policy is applied at the prompt, the tool call and the data edge. |
| Orchestration | Agent to agent handoffs are mediated with a verifiable trace and human in the loop hooks. |
| Audit | A replayable trail of agent decisions is kept for seven years. |
Behind these sits a proprietary corpus covering 280,000 SaaS vendors, four years of permission and scope history, and a reachability graph.
Identities And Systems Covered
| AI agents | Vendor-enabled agents and assistants including Claude, Copilot, n8n, Glean, Codex, Cursor, Ramp and Domo, plus IDE agents. |
|---|---|
| Non-human identities | Credentials and service accounts, with permissions, privileged access and authentication methods monitored. |
| SaaS identities | Shadow IT, OAuth token hygiene, account takeover and password spraying, from the SSPM heritage. |
| Systems | Okta, SailPoint, GitHub, Slack, Snowflake, Salesforce, Notion, Jira, Confluence and AWS. |
Company Background
| Headquarters | Tel Aviv, Israel. |
|---|---|
| Leadership | Galit Lubetzky Sharon, co-founder and CEO. |
| Funding | A reported $26 million across seed and Series A, from GGV Capital, Harmony Partners, S-Capital and Silicon Valley CISO Investments Group. |
Note: This overview is compiled from Wing 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.