Trent AI builds an agentic security platform that acts like an AI security engineer for software teams. It is meant for organisations shipping code and AI agents faster than their security staff can review them, and it looks for risks that can actually be exploited instead of producing long lists of alerts.
The platform reads code, infrastructure as code, configs, scripts, AI agents with their skills, tools and models, CI automations, connected tools and design documents. It runs continuously rather than as a one-off scan.
At A Glance
| Vendor | Trent AI Ltd |
|---|---|
| Category | Agentic AI security platform |
| Identities covered | AI agents, their skills, tools and models, plus code, cloud and CI automations |
| Deployment | Public cloud, customer VPC or on-premises |
| Agentic AI | Agent and attack chain analysis, coding agent plugins, hosted MCP server |
| Certifications | SOC 2 (Type I report on request) |
| Funding | Seed round, April 2026 |
| Market classification | Agentic AI Identity |
What Trent AI Does
Trent AI covers code, cloud and AI agents in three phases. First it builds understanding, reading code, infrastructure, configs and agents together so that it has a single picture of where real risk sits. Next it plans with the team, writing security rules mapped to the frameworks the organisation answers to and letting the team choose between autonomous operation and a human in the loop. Finally it secures, finding exploitable paths and fixing them with the team’s coding agents.
The output is organised as “Attack chains, not alert lists”, so engineers see how a weakness could be used end to end rather than a pile of isolated findings. Every fix is verified, and the whole process leaves a full audit trail.
Products
AI Security Engineer
The AI Security Engineer is the core product. It keeps its view of the environment current, judges which issues are exploitable, works out fixes with coding agents and confirms them afterwards. Teams use it in several ways:
- Tracking AI security posture over time by connecting repositories or agent definitions
- Securing AI applications, looking at agent behaviour, tool permissions, data flows and autonomous workflows
- Assessing a web application from its live URL, with or without source code
- Reviewing Claude Code and OpenClaw environments
AI Security Maturity Model (ASMM)
The ASMM is a public framework and self-assessment that helps organisations gauge how mature their security is for the agentic era.
Trentclaw
trentclaw is a skill that runs inside the third-party OpenClaw runtime, bringing Trent’s checks to OpenClaw users. It is self-serve on ClawHub and needs a Trent account and API key.
How It Works
Trent AI takes GitHub repositories, live web URLs, code uploads, design documents and specs, agent definitions written in Markdown, or even ideas before any code exists. Access is approval-gated.
| Component | What it does |
|---|---|
| Deployment | The platform runs in the public cloud, in the team’s own VPC or on-premises, and the team’s code and context stay inside its own boundary. |
| Hosted MCP | A hosted server listed in the MCP Registry as ai.trent/trent runs in US and EU regions, and data stays in the region the team signs in to. |
| Models | Teams can use frontier, open-source or private models. |
| Fix flow | Trent never merges code itself. The team’s coding agent prepares a pull request, Trent reviews it, the team or CI approves it, and Trent verifies the fix after merge. |
Identities And Systems Covered
| AI agents | Agent definitions, skills, tools and models, including agent behaviour, tool permissions and autonomous workflows. |
|---|---|
| Code and infrastructure | Application code, infrastructure as code, configs and scripts. |
| Automation | CI automations and the tools connected to them. |
| Web applications | Assessed from a live URL. |
Agentic AI Capabilities
Agent And Attack Chain Analysis
Analysis of agents targets prompt injection, tool misuse and unintended autonomous actions. Each attack chain is mapped to MITRE ATLAS and shows a confidence level, numbered steps, the controls already in place and the tasks that would break the chain. STRIDE, MITRE ATT&CK, MAESTRO and the OWASP Top 10 for Agentic Applications are also used.
Coding Agent Plugins
The Trent plugin for Claude Code and OpenAI Codex is open source under the MIT licence, and adds four Trent security skills to the coding agent. Other MCP clients connect over remote MCP with browser sign-in.
Integrations
| Area | Integrations |
|---|---|
| Coding agents and runtimes | Claude Code, OpenAI Codex, Cursor, Windsurf, Gemini CLI, Claude Desktop and Cowork, GitHub Copilot, OpenClaw |
| Security tools | Semgrep, Snyk and Sonar (native, read-only), SARIF upload from other tools |
| Team tools | Jira, Linear, Slack, Notion |
Company Background
| Founded | 2025, by Eno Thereska, Neil Lawrence and Zhenwen Dai. |
|---|---|
| Offices | London and San Francisco. |
| Leadership | Eno Thereska, Co-founder and CEO; Neil Lawrence, Co-founder and Chief Scientist; Zhenwen Dai, Co-founder and CTO. |
| Funding | Came out of stealth on 7 April 2026 with a $13M seed round led by LocalGlobe and Cambridge Innovation Capital, with participation from executives at OpenAI, Spotify, Databricks and AWS. |
| Recognition | The Hacker News 2026 Cybersecurity Stars Awards, agentic AI security platform category. |
Note: This overview is compiled from Trent AI’s publicly available information as of 4 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.