NOMINATION HIGHLIGHTS

Akto entered the market at the moment a new attack surface emerged and no security category existed to address it: the security of AI agents, systems that reason autonomously, chain tool calls, access sensitive resources, and execute actions across production environments.

Rather than extending existing tooling to cover agentic risk, Akto built a new security discipline from first principles, organized around three pillars that must all exist for an agentic security program to function: discovery and context mapping, adversarial red teaming, and runtime guardrail enforcement.

1. Discovery and context mapping. Akto discovers every AI agent, MCP server, connected tool, and data access path across 50+ connectors including shadow AI. The AI Agent Context Graph maps live behavioral relationships across agents, tools, prompts, permissions, memories, and action paths, making agentic risk visible in a form prior security tooling could not produce.

2. Adversarial red teaming. Agent Probe tests the way agentic systems actually fail: whether prompts trigger risky tool calls, whether agents can escalate beyond intended permissions, whether sensitive data is reachable through indirect action chaining, whether MCP-connected systems can be abused through orchestrated workflows. 4,000+ prebuilt test cases, CI/CD native, red teaming as a continuous control.

3. Runtime guardrail enforcement. This is where Akto’s deepest technical differentiation lives. The MCP Proxy sits inline between agent clients and MCP servers, enforcing before unsafe actions complete, not logging after the fact. The guardrail architecture covers five layers:
– Content and policy guardrails: prompt injection, context poisoning, intent violations
– Sensitive information controls: PII filtering with automatic anonymization, secrets detection
– Tool guardrails: tool misuse, malicious tool invocation, tool name-to-description mismatch
– Code execution controls: unexpected or unsafe code execution triggered by agent behavior
– Custom LLM-based policies: configurable to organizational risk thresholds

Every guardrail is independently configurable by server, direction, and enforcement mode (block or log). The built-in Playground validates every guardrail against real attack scenarios before deployment. Enforcement is proven before it goes live.

Akto is backed by Accel and Alumni Ventures. Investors and advisors include the founder of Tenable, the founder of Notion, the CEO of Sentry, Jim Manico, and the CTO of Synack. Recognized in Forbes, Dark Reading, VentureBeat, Nasdaq, and CSO Online.