Back to vendors
T

Trent AI

Visit site
Entry priceNot public; tailored proposal on requestFull pricing detail

AI security engineer whose four specialist agents scan code, cloud and AI agents, prioritize exploitable risk, open fix pull requests and verify them, reaching developers inside their coding tools through a hosted MCP server and plugin.

Trent AI is a London based security company founded in 2025. Chief executive Eno Thereska was a distinguished engineer at Alcion, which Veeam acquired, and earlier at AWS and Confluent; chief scientist Neil Lawrence is the DeepMind Professor of Machine Learning at the University of Cambridge and a former director of machine learning at Amazon; and chief technology officer Zhenwen Dai came from machine learning roles at AWS and Spotify. In April 2026 the company emerged from stealth with a thirteen million dollar seed led by LocalGlobe and Cambridge Innovation Capital.

Trent sells an AI security engineer that scans a customer's code, cloud and AI agents, finds exploitable risk, ships fixes and verifies them. Four specialist agents divide the work: a Threat Scanning Agent observes agents, code, infrastructure and dependencies; an Analysis Agent separates signal from noise and prioritizes by business impact; a Remediation Agent patches vulnerabilities, opens pull requests and adjusts configurations; and a Security Posture Agent tracks trends and benchmarks against standards such as SOC 2 and NIST.

Since August 2026 the work is organized as Understand, Plan and Secure, with every finding linked to its evidence and severities that users can challenge and regrade in place. Trent reaches developers inside Claude Code, Codex, Cursor, GitHub Copilot and other coding tools through its hosted MCP server, now delivered as a plugin with US and EU endpoints, and in September 2026 it added read only Security Connectors that pull findings from Semgrep, SARIF tools, Notion and Linear into one judgment.

Trent states it can run in public cloud, the customer's VPC or on premises, with frontier, open source or private models. Organization workspaces carry Admin and Member roles, and a trust center sits at trust.trent.ai. Pull requests go through GitHub's own merge gate instead of an approval step inside Trent, and the MCP server's tool list is not published.

Pricing is tailored by segment, from solo builders to regulated enterprises, through a proposal. It fits an engineering or security team that wants security review embedded in its coding tools and repositories across code, cloud and its own agents. It is early stage, with no published tool catalog, no attestation named outside its trust center and a short operating record.

Vendor details

Canonical URL

https://trent.ai

Category

Security / SOC agent

Subcategory

Security for AI agents

Funding status

Independent, headquartered in London, founded in 2025 by Eno Thereska (CEO, formerly a distinguished engineer at Alcion, AWS, and Confluent), Neil Lawrence (chief scientist, DeepMind Professor of Machine Learning at Cambridge and former Amazon ML director), and Zhenwen Dai (CTO, formerly of AWS and Spotify). Trent emerged from stealth in April 2026 with a thirteen million dollar seed led by LocalGlobe and Cambridge Innovation Capital, with angels including OpenAI's Joaquin Quinonero Candela, AWS director Avinash Bhat, Databricks engineer Ippokratis Pandis, and former Spotify AI leader Tony Jebara. Early design partners include Canopy, Commscentre, ML@Cam, Qbeast, and Weblogic.

Company status

independent

Use cases & customers

Primary use cases

Securing enterprise AI agentsDetecting prompt injection and tool misuseAutonomous vulnerability remediation for agentsAgent security posture management

Target customers

Engineering teams building AI agentsEnterprise security teamsDevelopers of autonomous workflowsAI platform and governance teams

Deployment options

Cloud

Integrations

Trent connects to repositories through a GitHub App, where its Remediation Agent opens pull requests, takes code uploads and live application URLs, and reaches developers in Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Lovable and Claude Desktop through its hosted MCP server, delivered as a plugin, and runs as a skill in OpenClaw. Security Connectors, launched in September 2026, read findings from Semgrep, SARIF compatible scanners, Snyk and Semgrep JSON, Notion and Linear and write nothing back.

In practice

A team ships an AI agent with access to real systems and no security review built for it. Trent's Threat Scanning Agent maps where risk lives across the code, infrastructure and agents, and its Analysis Agent prioritizes what is exploitable.

A repository carries a vulnerability that could let an agent be steered into exfiltrating data. Trent's Remediation Agent opens a fix pull request or hands a ready to run prompt to the developer's coding assistant, then verifies the fix landed.

Security wants to keep pace with builders without slowing them down. Trent runs inside Claude Code, Cursor and other coding tools through its MCP plugin, and its Security Connectors bring existing scanner findings into one prioritized view.

Agentic Index coverage score

8.0 / 14 capabilities · 57%

Integrations & Tool Calling Partial

The Remediation Agent opens pull requests and adjusts configurations through a GitHub App connection, and fixes can also go to the customer's coding assistant as a ready to run prompt. GitHub is the only write connector named. The Security Connectors pull findings from Semgrep, Notion, Linear and SARIF tools and are read only, so Trent reads those sources and writes nothing back.

Sourcetrent.ai/productread 2026-09-28

Workflow Orchestration Full

Four specialist agents divide the work. They are a Threat Scanning Agent, an Analysis Agent that classifies signal from noise and prioritizes by business impact, a Remediation Agent and a Security Posture Agent.

Sourcetrent.ai/productread 2026-09-28

Knowledge Grounding & RAG Full

A context layer built by Trent covers the customer's application code, infrastructure code, configurations, agents and automations. Its Inventory and Architecture Review maps software, services, agents and data flows, later rescans keep that map current, and the agents reason over it. Trent does not say how the context layer is stored.

Sourcetrent.airead 2026-09-28

Human Oversight & Guardrails Partial

Users can challenge and regrade a severity in place, edit draft requirements and make task decisions. Trent pitches the product as "Built to Be Corrected, Not Blindly Trusted", and fixes arrive as pull requests or prompts for the developer, so the decision stays with the team. The merge gate belongs to GitHub, and Trent names no approval step of its own that holds an agent action.

Sourcetrent.ai/productread 2026-09-28

Security, Identity & Governance Partial

Organization workspaces carry role based access control. Admins manage members and roles, create GitHub App connections, and view and revoke any member's API keys. Members run analyses and plan remediation. A trust center sits at trust.trent.ai, and no attestation is named outside it.

Sourcetrent.ai/blog/organization-workspaces-role-based-access-controlread 2026-09-28

Observability & Auditability Partial

Every finding links back to its evidence with an explanation of how Trent graded it. A Report changelog shows what moved after each rescan, and fixes are confirmed with an audit trail. Together these explain the agents' verdicts, but Trent names no run trace of the agents' own steps. Posture tracking reports on the customer's estate, not on the agents.

Sourcetrent.ai/productread 2026-09-28

Memory & State Persistence Not documented

The scanning agent, by Trent's account, learns where to look and reduces noise over time. The context layer holds knowledge about the customer's estate, and Trent names no agent memory with a stated scope and lifetime.

Sourcetrent.ai/productread 2026-09-28

Deployment & Data Residency Full

Customers can run Trent in public cloud, in their own VPC or on premises, wherever their security and compliance posture demands. The hosted MCP service has separate US and EU endpoints at mcp.trent.ai and mcp.eu.trent.ai.

Sourcetrent.airead 2026-09-28

Prebuilt Agents, Templates & Packs Full

Four prebuilt agents ship with Trent, each with its own job. They are Threat Scanning, Analysis, Remediation and Security Posture. A plugin for Claude Code and Codex CLI packages review, threat modeling and remediation tracking skills.

Sourcetrent.ai/productread 2026-09-28

Triggers & Channel Coverage Partial

Scanning of code, cloud and agents is continuous by Trent's account, and rescans update the findings. No event, webhook or schedule that starts the agents is named, and the plugin and MCP tools run when the developer's assistant calls them.

Sourcetrent.ai/productread 2026-09-28

Model Flexibility & Routing Full

Customers can run Trent with frontier, open source or private models, including their own. No provider list is published.

Sourcetrent.airead 2026-09-28

APIs, SDKs & MCP Extensibility Partial

A hosted MCP server of Trent's own runs at https://mcp.trent.ai/mcp and https://mcp.eu.trent.ai/mcp. Users sign in through the browser, or use one API key for a remote connection. The server comes through an MIT licensed plugin for Claude Code and Codex CLI. Its tool list is not published, and Trent names no REST API or SDK.

Sourcegithub.com/trnt-ai/trent-agent-pluginread 2026-09-28

Testing, Debugging & Optimization Not documented

The Remediation Agent validates that fixes work, and the Posture Agent benchmarks against standards. Both test the customer's code and estate, not Trent's agents. Trent names no harness, scored test cases or quality gate for the agents.

Sourcetrent.ai/productread 2026-09-28

Browser & Computer Use Not documented

Trent reads repositories, uploaded code and live applications by URL, and works through MCP and GitHub. It names no browser or desktop operation by an agent.

Sourcetrent.ai/integrationsread 2026-09-28

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Product updates

·Agent capabilityVerified

Version 0.3.0 of Trent's plugin for Claude Code and Codex adds a trent-loop skill. After a scan, the coding agent fixes up to 10 open findings on a new branch, checks each fix with the repo's tests and Trent's security advisor, and opens one pull request. The same release drops the step where a user signed off on the remediation plan before fixes started.

Bears on: Human approval / guardrails

View source
·Human approval / guardrailsVerified

Trent AI introduced a new interactive experience for its AI Security Engineer, enabling users to review the agent's work incrementally instead of waiting for a complete scan. The update organizes workflows into a three-part structure: Understand, Plan, and Secure. It also allows users to trace findings back to their evidence, follow attack paths, and regrade severities in place.

Bears on: Human approval / guardrails

View source
View all 2 Trent AI updates →Tracked since Aug 2026 · Verified from public vendor sources

Pricing

Not public; tailored proposal on request

tailored proposal; basis not disclosed

What is public

A pricing page with no prices. It says pricing is tailored to the team and stack, names three segments (Solo Builders and Startups, Engineering Teams, Enterprise and Regulated Industries) and offers Request Pricing.

Billing mechanics

Proposal based; the pricing page names three segments but no tiers, units or billing period.

Cost watchouts

Inference, not stated by the vendor: scope may grow with the repositories, applications and agents covered.

Variable cost rationale

Inference, not stated by the vendor: with tailored pricing and no published unit, cost likely scales with the code, cloud and agents in scope.

Additional watchouts

Ask how the segments differ and what unit the proposal prices on (repositories, developers or agents), and confirm terms given the company's early stage.

Sales call required

Yes, required for paid access

Free / trial

No free tier or trial published; request pricing

Key ambiguities

No public rate, entry point or scope unit; the three segments on the pricing page carry no prices.

Agentic Index verified 2026-09-28

Alternatives to Trent AI

The closest documented capability profiles to Trent AI among security and SOC agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

  • Crogl9.0 / 14Fuller documented coverage on Integrations & Tool Calling and Observability & Auditability
  • ThreatModeler10.0 / 14Adds documented Testing, Debugging & Optimization
  • Seemplicity8.5 / 14Fuller documented coverage on Integrations & Tool Calling and Triggers & Channel Coverage
  • UnderDefense8.5 / 14Fuller documented coverage on Integrations & Tool Calling and Observability & Auditability
  • Airrived10.0 / 14Adds documented Testing, Debugging & Optimization
  • Enclave7.0 / 14Fuller documented coverage on Integrations & Tool Calling and Triggers & Channel Coverage

Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.