Salt Security protects the APIs that AI agents and MCP servers run on, and now the agents themselves. It is built for security teams that need to know what their agents can reach, fix the paths that create risk, and stop abuse at the point where an agent takes an action.
Everything sits in its Agentic Security Platform, which looks at agents, LLMs, MCP servers and APIs from four angles: external exposure, source code, configuration and runtime behaviour.
At A Glance
| Vendor | Salt Security |
|---|---|
| Category | API and agentic AI security |
| Identities covered | AI agents, MCP servers, APIs, hard-coded credentials in code |
| Deployment | Agentless discovery, traffic collection, enforcement through existing gateways and WAFs |
| Agentic AI | MCP discovery, policy inside AI coding agents, runtime protection |
| Customers | Scotiabank, Luxottica, Standard Bank Group, Hyundai, SoFi |
| Market classification | Application Security; also Agentic AI Identity |
What Salt Does
AI agents and MCP servers run on APIs, so an agent’s risk is the risk of the APIs and MCP servers it can reach. Salt maps every agent, MCP server, gateway and app into one graph and carries context along the same path an agent’s request takes: from code before deployment, to the prompt, to the action. Posture and compliance checks sit across the whole graph so teams can catch configuration drift and correct it quickly. Enforcement then happens at the moment an agent acts, not only in a later report.
Products
Salt Surface
Salt Surface answers the question most teams cannot: which APIs, MCP servers and agent endpoints are reachable from the internet, and which of them have no authentication. Starting from a list of domains, it rebuilds the exposure picture from mobile apps, front-end JavaScript, developer documentation and threat intelligence, including servers nobody registered, such as those inherited through acquisitions or left behind by old projects.
Salt Code
Salt Code brings security policy into the AI coding assistants developers already use, turning policies into rules that guide the assistant as it writes code. It works with Claude Code, Cursor, GitHub Copilot, Windsurf, Kiro, Codex, Gemini CLI, VS Code, JetBrains and any MCP client. More than 140 pre-built policies ship with it, teams can add their own, and the same checks extend into CI/CD and pull requests so violations are stopped before production.
Salt Connect And Salt Collect
Salt Connect reads the API inventory in a customer’s cloud environments, and Salt Collect analyses live traffic. Between them they supply the inventory and runtime data that Surface, Code and Protect depend on.
Salt Protect
Salt Protect stops attacks that abuse an API’s logic rather than exploit a known signature. It builds a baseline from live traffic for each API, user, session and agent, and uses “patented behavioral analysis” of sequences over time to spot intent. Alerts go to SIEM and SOAR tools.
How It Works
| Discovery | Agentless to start, with Surface mapping external exposure from outside and Connect pulling API inventory from the customer’s cloud environments. |
|---|---|
| Traffic analysis | Collect feeds live traffic into the behavioural baselines. |
| Enforcement | Protect can be added later without rerouting traffic or replacing existing controls, because blocking happens in the customer’s own gateways and WAFs. |
| Code and pipeline | Code applies one policy model from the coding assistant through CI/CD to production configuration. |
Identities And Systems Covered
| AI agents | Per-agent baselines and a map of which APIs each agent can reach |
|---|---|
| MCP servers | Exposed and unregistered MCP servers, with authentication checks |
| APIs and users | Per-API, per-user and per-session baselines from live traffic |
| Credentials in code | Hard-coded keys flagged by source code scanning |
| Platforms | AWS, Microsoft Azure, Google Cloud Platform, Kubernetes and Docker |
Agentic AI Capabilities
| MCP discovery for Claude Enterprise | Salt Claude Connect gives continuous visibility into the MCP servers connected to Claude Enterprise. |
|---|---|
| Action layer | Through its GitHub integration, Salt covers the actions agents take, not only the APIs they call. |
| LLM runtime and prompt security | The platform watches model runtime and prompts alongside API traffic. |
Integrations
| Area | Integrations |
|---|---|
| AI and code platforms | Claude Enterprise, GitHub |
| Gateways and edge | Kong, Apigee, MuleSoft, Cloudflare, Akamai, F5 |
| Security and SOC | CrowdStrike, Microsoft Sentinel, Wiz, Veracode |
Company Background
| Founded | 2018, by Roey Eliyahu and Michael Nicosia |
|---|---|
| Headquarters | Palo Alto, California, United States |
| Leadership | Roey Eliyahu (Co-founder and CEO), Michael Nicosia (Co-founder and COO), Michael Callahan (Chief Marketing Officer), Kfir Lippmann (Chief Financial Officer) |
| Investors | Advent International, Alkeon Capital Management, CapitalG, DFJ Growth, S Capital VC, Sequoia, Tenaya Capital and Y Combinator |
| Scale | More than 100 reported enterprise deployments and 40+ technology integrations |
| Research | Salt Labs, the company’s research team, publishes vulnerability research and a State of AI and API Security report |
| Recognition | Three InfoSec Innovator Awards for 2026 |
Note: This overview is compiled from Salt 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.