“Agents are cool. Attacks are not.” That’s the philosophy of Ankita Gupta, the CEO, as she recently introduced the new Agentic Security Platform from Akto. Their Agentic Security Platform is an end-to-end security solution purpose-built for the age of autonomous AI agents. Designed to protect enterprises as they deploy AI agents at scale and adopt the Model Context Protocol (MCPs), the platform delivers complete visibility, continuous Agentic red teaming, and real-time guardrails to stop agents from going rogue.
As AI agents like ChatGPT, Claude, and GitHub Copilot rapidly become part of core enterprise workflows, developers are adopting MCP to enable these agents to safely and dynamically invoke APIs. But with this shift comes a new attack surface – MCP Servers. For example, in May 2025, a critical vulnerability in GitHub’s MCP server allowed attackers to embed malicious instructions in public issues. When processed by AI agents, those instructions led to unauthorized access and data leakage highlighting the urgent need for MCP-specific security controls.
“MCP is the protocol powering the next generation of AI-native software,” said Ankita Gupta, Co-Founder and CEO of Akto. “The risks introduced by agents are unlike anything enterprises have faced before. Prompt injection can cause cascading hallucinations. Rogue or malicious tools can misuse enterprise privileges. Memory can be poisoned and data sources manipulated. Autonomous agents can execute financial or operational transactions without proper oversight. Traditional AppSec frameworks were not built to manage Agentic Risks. What enterprises need is a new security discipline designed for this agentic era – guardrails designed for Agents, MCP servers, tools, resources. So to help organizations manage and secure this new discipline, we’re beyond excited to launch Akto’s Agentic Security Platform.”
Akto MCP Security is designed from the ground up to protect MCPs. It detects shadow MCP servers, tests for prompt injection and tool poisoning vulnerabilities, and monitors AI-to-API traffic in real time to flag suspicious behavior. The platform helps security teams stay ahead of threats in a world where APIs aren’t just passive endpoints, they’re actively invoked by autonomous agents that can introduce new risks with every interaction.
“Our platform secures every layer of the agentic ecosystem, covering agents, MCP servers, tools, prompts, and resources,” Gupta stated. Built in collaboration with Akto’s enterprise customers, the Akto MCP Security platform includes three core modules at launch.
MCP Server Discovery automatically detects all MCP-compatible servers and related API calls across environments using Akto’s 50+ traffic and code connectors, eliminating shadow MCPs. Agentic Visibility & Discovery:
MCP Security Testing continuously tests MCP endpoints and tools for vulnerabilities like unauthorized access, prompt injection, insecure auth, and data exposure.
MCP Monitoring & Threat Detection provides real-time behavioral analysis of MCP traffic to detect threats such as tool misuse, permission escalation, and malicious agent behavior.
Akto’s Agentic Security Platform is also built on three core pillars, which secures every layer of the agentic ecosystem, covering agents, MCP servers, tools, prompts, and resources.
“The first pillar is Agentic Visibility and Discovery, which provides a complete inventory of all agents, MCP servers, tools, and data sources, while also mapping their lineage to track dependencies and risks,” Gupta said. “The second pillar is Agentic Red Teaming, where Akto continuously tests agentic assets with 1000+ AI probes to uncover prompt injection attempts, poisoned resources, and tool misuse before attackers can exploit them. The third pillar is Guardrails and Runtime Protection, which enforces enterprise policies to restrict risky or unauthorized actions and blocks cascading failures in real time. These capabilities combine to deliver an integrated Agentic security posture. With Akto, enterprises gain unified visibility across every agent, MCP server, tool, resource, and prompt.”
“MCP opens powerful new possibilities, but also dangerous new paths for abuse,” said Ankush Jain, CTO at Akto. “We’ve built dedicated logic for how MCP works, so security teams can monitor, test, and protect these systems with context.”
So who is this for?
“Agentic Security is built for Modern AI Security teams and enterprises building and adopting Agents, MCPs across critical business functions,” Gupta indicated. “Wherever agents are deployed, security and governance must follow.”
She noted that Akto was the first-mover in MCP Security.
“Akto was the first to launch MCP Security. We’ve been shaping this space since June 2025, and are now extending it into full Agentic Security platform. We have the world’s largest AI Agent Attack Matrix, with 1,000+ documented probes and attacks across agents, prompt, tool, memory, and orchestration layers. We don’t just talk about risk. We weapon-test it like attackers and deliver guardrails that enterprises trust.
“The shift to agentic systems is as transformative as the move from monolithic apps to APIs a decade ago or from on-prem to cloud 15 years ago,” Gupta concluded. “Teams that addressed the risks early set the standard for the industry. The same will hold true for agents today.”