Operant AI is a runtime security company that protects AI applications, AI agents, MCP servers and the APIs and Kubernetes workloads beneath them. Its customers are enterprises that are putting agents into production and need to see, and control, what those agents and tools actually do.
The platform watches live traffic and behaviour rather than reviewing configurations after the fact. It runs from cloud workloads to employee laptops, where coding agents and unsanctioned AI tools increasingly operate, and it treats agents and MCP connections as non-human identities that need their own access controls.
At A Glance
| Vendor | Operant AI, Inc. |
|---|---|
| Category | Runtime security for AI agents, MCP, APIs and cloud workloads |
| Identities covered | AI agents and agentic non-human identities, MCP clients and servers, APIs, employee devices |
| Deployment | Kubernetes-native and hybrid cloud, private SaaS, VPC, on-premises and air-gapped; endpoint app for macOS, Windows and Linux |
| Agentic AI | Agent Protector, MCP Gateway, Semantic Firewall, Endpoint Protector |
| Certifications | SOC 2 Type II |
| Funding | Series A led by SineWave Ventures and Felicis (2024) |
| Market classification | Agentic AI Identity; also Workload Identity |
What Operant AI Does
Agents call tools, tools call APIs, and a weakness anywhere in that chain can expose data. Operant’s approach is to secure the whole chain at runtime rather than one link of it, starting with the APIs and Kubernetes workloads underneath and working up through AI applications, MCP traffic and the agents themselves to the employee workstation. Each layer is enforced where the traffic actually flows, and policy decisions increasingly turn on what an action is trying to achieve rather than on fixed rules about endpoints.
Products
API & Cloud Protector
API & Cloud Protector is for teams whose APIs run on Kubernetes-native stacks and need threat protection beyond a web application firewall. It covers both north-south traffic entering the cluster and east-west traffic between services.
AI Gatekeeper
AI Gatekeeper defends AI applications and agents at runtime. It detects threats against agents, assigns them trust scores and blocks unauthorised ones, applies OWASP-aligned threat modelling and extends identity and access controls to MCP and AI non-human identities. It works without the cost or overhead of VPC traffic mirroring.
MCP Gateway
MCP Gateway, part of AI Gatekeeper, gives security teams control over the MCP tools their agents use. It builds live catalogues of MCP tools and agents across environments, applies least-privilege execution and granular permissions to tool usage, and sets trust zones that block untrusted servers and tools, with rate limiting and encryption enforcement. A central governance framework covers agent and tool management across the enterprise.
Agent Protector
Agent Protector addresses the agents that appear across cloud and SaaS faster than security teams can track them. It discovers unmanaged agents, MCP servers, tools and their dependencies, builds agentic identity catalogues and enforces zero trust controls designed for agentic AI.
Endpoint Protector
Endpoint Protector covers the AI tools employees run on their own machines. It inventories sanctioned and unsanctioned AI tools, MCP servers and plugins, watches for prompt injection, identity anomalies and exfiltration of personal data, and redacts secrets and keys automatically.
Semantic Firewall
Semantic Firewall enforces policy on intent. It works out the purpose behind every prompt, tool call, command and data movement in real time and returns allow, block or redact decisions. Four guards split the work:
| Tool Intent Guard | Blocks exfiltration, bulk dumps, credential access and unauthorised sharing |
|---|---|
| Code Intent Guard | Stops malicious execution, injection, shell breakout and privilege escalation by coding agents |
| Data Intent Guard | Applies Microsoft Purview sensitivity labels |
| Scope Guard | Keeps agents within scope, with natural-language policies such as “no unauthorized deletes” |
Woodpecker
Woodpecker is an open-source automated red teaming engine. It tests Kubernetes misconfigurations, API security and AI models for prompt injection and data poisoning.
How It Works
The cloud products are Kubernetes-native and run from Kubernetes to hybrid and private clouds, including remote agents on Google Vertex AI.
| Component | What it does |
|---|---|
| Endpoint layer | Endpoint Protector is a native workstation application, rolled out across fleets through standard MDM and JAMF workflows and tied to the enterprise identity provider. |
| Data handling | In private mode, sensitive prompts, agent traces and detected data stay inside the customer’s environment, and private SaaS options support data residency. |
| Decision point | Semantic Firewall classifies intent with Operant’s own models, and every decision is made inside the enterprise environment, including VPC, on-premises and air-gapped deployments. |
Identities And Systems Covered
| AI agents | Managed and unmanaged agents. |
|---|---|
| Agentic non-human identities | Agent and MCP identities, with identity and access controls. |
| MCP | Servers, clients, tools and connections. |
| Secrets | Secrets and keys on employee endpoints. |
| Cloud platforms | Kubernetes, Amazon EKS, Fargate, Bedrock, Azure and Google Cloud. |
| Data and SaaS platforms | Databricks, Snowflake and Salesforce. |
| LLM providers | OpenAI, Deepseek, Cohere, Anthropic and Hugging Face. |
Company Background
| Founders and leadership | Vrajesh Bhavsar (co-founder and CEO), Dr. Priyanka Tembey (co-founder and CTO) and Ashley Roof (co-founder and CMO). |
|---|---|
| Headquarters | San Francisco, United States. |
| Funding | $10M Series A in 2024. |
| Recognition | Featured by Gartner across five AI security categories, including the AI TRiSM Market Guide and the API Protection Market Guide. |
Note: This overview is compiled from Operant AI’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.