What Trent AI does

Trent AI Limited is a London-based, venture-backed startup founded in 2025 that builds an agentic AI security platform branded the "AI Augmented Security Engineer." Its core product is a continuous loop of four specialised agents (Scan, Judge, Mitigate, Evaluate) orchestrated by a proprietary reinforcement-learning-trained judgment layer, which delivers context-aware security assessments for LLMs, AI applications, autonomous agent systems and traditional web apps. The platform is exposed via Model Context Protocol servers directly inside Claude Code and Lovable, and as an open-source skill (trentclaw) inside OpenClaw, and is extended by solution modules for AI Security Posture Management, AI Application Security, URL-based Web Application Security Assessment, and the AI Security Maturity Model (ASMM) - a 28-category framework mapped to NIST CSF 2.0, NIST AI RMF, the EU AI Act and OWASP Top 10 for Agentic Applications 2026.

Trent serves three customer segments defined on its pricing page - Solo Builders & Startups, Engineering Teams, and Enterprise & Regulated Industries - through a hybrid sales-led and PLG motion. Commercial revenue comes from custom-quote annual subscriptions delivered via direct enterprise sales (with SOC 2 displayed on the Contact page), supported by a free OpenClaw API-key funnel into trentclaw and bespoke ASMM engagements. Named customers and design partners include Arvest Bank, Canopy, ML@CL, weblogic, Qbeast and CommsCentre. Pricing is not publicly disclosed; free trials convert automatically to paid subscriptions unless cancelled before expiry.

Trent closed a $13M seed round in April 2026 co-led by LocalGlobe and Cambridge Innovation Capital, with participation from executives at OpenAI, Spotify, Databricks and AWS, and emerged from stealth alongside the launch of seven product surfaces. The company operates with 1-10 employees, is led by founders Eno Thereska (ex-Distinguished Engineer at AWS/Confluent/Alcion), Neil Lawrence (DeepMind Professor of ML at Cambridge) and Zhenwen Dai (former head of RL at Spotify and Amazon), and was named Agentic AI Security Platform Leader in the 2026 Cybersecurity Stars Awards by The Hacker News in June 2026.

Name

Trent AI

Legal name

Trent AI Limited

Website

https://trent.ai

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

11–50 employees

Short description

Trent AI builds an agentic AI security platform that continuously scans, judges, mitigates and evaluates risk across LLM applications, autonomous agents and web apps, sold to enterprise security teams, engineering teams and solo builders via custom subscriptions and an open-source OpenClaw skill.

Ownership category

akta.pro rank

Headquarters

HQ city

London

HQ country

United Kingdom

HQ region

Europe

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Marketing or Sales, Operations, Infrastructure

Revenue model

1. Subscription / platform licensing (enterprise and team plans): Recurring paid subscriptions segmented into tiers for Solo Builders & Startups, Engineering Teams, and Enterprise & Regulated Industries. Pricing is not publicly disclosed; customers are onboarded via direct sales with custom proposals. Free trials convert automatically into paid-for subscriptions unless cancelled before expiry per the Terms of Service.
2. Freemium OpenClaw skill (trentclaw) with API-key access: Free self-serve access to the OpenClaw security assessment skill via a Trent API key, intended as a community/PLG funnel into the broader commercial Trent platform.
3. Professional services / managed security engagements: Trent performs bespoke enterprise customer engagements (e.g., contracted assessments, ASMM maturity-model delivery) per Online Orders under the Terms of Service; ASMM requests are explicitly reviewed before manual delivery rather than auto-sent.

Pricing tiers

Go-to-market motion1 record

Distribution channels5 records

Marketing channels8 records

Core offering

Trent AI builds an agentic AI security platform that delivers continuous, automatic protection for LLMs, AI applications, and autonomous agent systems. The product is structured around four compounding agents — Scan, Judge, Mitigate, and Evaluate — orchestrated by a proprietary reinforcement-learning judgment layer, and is delivered via native MCP integrations into developer environments (Claude Code, Lovable, OpenClaw) plus a request-based AI Security Maturity Model for enterprise programs.

Product overview

Trent AI offers a single, unified agentic AI security platform — branded the "AI Augmented Security Engineer" — composed of four specialized agents (Scan, Judge, Mitigate, Evaluate) that work in one continuous compounding loop, rather than a patchwork of disconnected point tools. Layered on top of this core engine are solution modules targeting different surfaces: AI Security Posture Management (AI-SPM) for cross-portfolio AI risk, AI Application Security for specific AI apps, and Web Application Security Assessment for URL-based scans of vibe-coded apps. Trent also ships dedicated integration modules for popular AI development surfaces: Claude Code Security (via Trent's MCP server in the Claude Code IDE), Lovable Security (via MCP into Lovable projects), and OpenClaw Security (installed as the open-source trentclaw skill in OpenClaw). The portfolio is rounded out by the AI Security Maturity Model (ASMM) — a 28-category, 6-domain NIST CSF 2.0-aligned assessment framework delivered as a request-based maturity assessment for enterprise security leaders. Trent was recognized as the 2026 Agentic AI Security Platform Leader in The Hacker News' Cybersecurity Stars Awards.

Differentiator

Problem solved

Functional benefit

Brands

• trentclaw: Trent AI's open-source OpenClaw Security Assessment Skill that runs continuous security audits on OpenClaw configurations, skills, and tool permissions; distributed via ClawHub.
• AI Security Maturity Model (ASMM)

Products and services

• Trent AI Platform (AI Augmented Security Engineer) Unified agentic AI security platform composed of four specialized agents — Threat Scanning, Analysis (Judge), Remediation (Mitigate), and Security Posture (Evaluate) — that work in a continuous self-reinforcing loop. Each cycle through the loop makes the next smarter, compounding security intelligence alongside the customer's AI agents and applications.
• AI Security Posture Management (AI-SPM) Continuous AI Security Posture Management for AI agents, LLM-powered applications, and agentic workflows. Trent's specialized agents continuously scan, judge, mitigate, and evaluate AI risk across the customer's environment, addressing prompt injection, tool misuse, data exfiltration through agent chains, and privilege escalation.
• AI Application Security Security assessment for specific AI applications, focused on the new threat surfaces created when an AI application reasons, plans, and acts autonomously — including function-call tool misuse, prompt injection, and unintended autonomous actions.
• Web Application Security Assessment External security assessment of a web application from just a URL — no source code access required. Trent's agents map exposed endpoints, authentication flows, headers, and configurations, then deliver a prioritized fix plan targeted at vibe-coded applications built with Lovable, Bolt, Replit, or Base44.
• Claude Code Security Trent's MCP server installed into Claude Code so security assessments run inside the IDE and the prioritized remediation plan feeds into Claude Code's task system. Continuously re-assesses as the codebase evolves.
• Lovable Security Connects Trent to a user's Lovable project via MCP and GitHub to scan the generated app, build a prioritized mitigation plan, and apply fixes directly inside Lovable without context switching.
• OpenClaw Security Continuous visibility across the OpenClaw agent runtime. Trent installs as a skill that audits configuration, environment variables, installed skills (including custom code), gateway exposure, and tool permissions for the OpenClaw environment.
• trentclaw (OpenClaw Security Assessment Skill) Open-source OpenClaw skill that performs a continuous security audit of OpenClaw configuration, custom code, and installed skills. Distributed via ClawHub; source on GitHub (github.com/trnt-ai/trent-openclaw-security-assessment). Benchmark study showed 94.6% recall on dangerous skills.
• AI Security Maturity Model (ASMM) A standards-aligned maturity framework that scores an organization's AI security program across 28 categories and 6 NIST CSF 2.0 domains (Govern, Identify, Protect, Detect, Respond, Recover). Produces a prioritized roadmap aligned to NIST CSF 2.0, NIST AI RMF 1.0, CSA AI Controls Matrix, EU AI Act, and OWASP Top 10 for Agentic Applications (2026). Delivered on a request basis.

Named customers6 records

Segments1 record

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration6 records

AI capability11 records

Feature7 records

Partnerships

Six partnerships are on record, tiered flagship integration, flagship integration / distribution channel, core integration and core ecosystem membership.

• Anthropic (Claude Code)flagship integrationTechnology or IntegrationTrent's MCP server integrates directly into Anthropic's Claude Code IDE; security assessments run in-IDE and the remediation plan feeds into Claude Code's task system. The "Your code never goes to Anthropic or OpenAI" claim is explicitly made on the homepage.
• Lovableflagship integrationTechnology or IntegrationTrent connects to Lovable via its MCP server; security fixes are applied directly inside Lovable once the user approves Trent's mitigation plan. Designed for the vibe-coding developer persona.
• OpenClawflagship integration / distribution channelTechnology or IntegrationTrent's open-source trentclaw skill is installed inside the OpenClaw agent runtime and audits configuration, environment variables, and installed skills. OpenClaw also serves as Trent's distribution surface via ClawHub.
• GitHubcore integrationTechnology or IntegrationTrent integrates with GitHub repositories for code analysis and applies fixes directly through repositories connected to Claude Code, Lovable, and OpenClaw projects.
• OWASPcore ecosystem membershipOthersListed as "Proud Members Of" on the Trent AI homepage alongside Carnegie Mellon CyLab. Trent's ASMM and product capabilities explicitly map to OWASP Top 10 for Agentic Applications (2026) and OWASP Top 10 for LLM Applications.
• CyLab Security and Privacy Institute, Carnegie Mellon UniversityOthers

Scale indicators9 records

Recent moves6 records

Expansion highlights6 records

Direct peers

• Protect AI: AI/ML security platform offering model scanning, supply-chain security, and LLM guardrails. Competes with Trent across AI application security posture and vulnerability remediation for AI systems.
• Calypso AI: AI security and governance platform for generative AI and LLM applications, with validation, red-teaming, and policy enforcement. Overlaps with Trent's AI Application Security and ASMM-style governance offerings.
• Robust Intelligence (Cisco): Acquired by Cisco, Robust Intelligence offers AI firewall and model security capabilities. A more mature competitor in AI application and model protection that Trent must differentiate against, particularly in enterprise.
• PromptArmor: AI security startup focused on prompt injection and agent misuse. Comparable product surface to Trent's AI Application Security and OpenClaw/Claude Code security integrations.
• HiddenLayer: AI security company providing adversarial threat protection and model security. Directly comparable in offering model and AI application risk assessment to enterprise security teams.
• Lakera: Direct competitor in LLM/agent security, focused on prompt injection and AI application guardrails. Targets similar enterprise and developer buyers and overlaps with Trent's AI Application Security and AI-SPM modules.
• Noma Security: AI security and governance platform targeting enterprise AI applications and agentic workflows. Comparable customer base (regulated enterprises) and similar AI risk posture management positioning.
• Cranium: AI security and compliance platform focused on enterprise AI systems. Overlaps with Trent's AI-SPM and ASMM capabilities, and addresses similar threat-surface expansion from generative AI.

Broad incumbents

• Snyk: Broad incumbent in developer security (SAST, SCA, IaC). Competes less directly than pure-play AI security peers but overlaps with Trent's URL/web application security and code-level vulnerability remediation.
• GitGuardian: Code and secrets security platform with AI-asset discovery capabilities. Adjacent to Trent's continuous scanning posture across source repositories and overlaps in enterprise security buyer personas.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat6 records

Key risks5 records

Key highlights6 records

Customer concentration

• GitHubgithub.com/trnt-ai/trent-openclaw-security-assessment
• Xx.com/TrentAIHQ
• LinkedInlinkedin.com/company/trentai

Compliance3 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles7 records

Funding overview

Funding rounds1 record

Investors2 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Trent AI