# Trident

> Trident is an AI-powered security platform for continuous web and API penetration testing and cloud attack-path analysis. It connects cloud assets, identities, data, access relationships, and application findings so teams can prioritize risks that may reach sensitive data.

Key facts:

- Product: cloud attack-path analysis, web and API penetration testing, and pull request security review.
- Cloud scope: maps assets, identities, data stores, exposure, and access relationships across AWS, Azure, and Google Cloud.
- Application scope: tests web applications and APIs and records findings for security review.
- Pull request review: trident-sentinel, Trident's GitHub App, reviews each pull request when it opens and on every push, comments on the changed lines, and reports a "Trident security review" check.
- Where teams use it: mention @Trident in Slack to ask about findings, scans and pull request reviews; coding agents such as Claude Code, Cursor and VS Code Copilot connect through a remote MCP server.
- Testing model: uses AI-assisted, continuous security testing while preserving authorization, reproducible evidence, human review, and retesting.
- Prioritization: combines cloud context and application findings to help teams focus on risks that could reach sensitive data.
- Company: Trident is operated by Esprit Labs Inc.
- Assurance: Esprit Labs Inc. holds a SOC 2 report issued by the independent audit firm Sensiba LLP, and monitors its controls continuously with Vanta. SOC 2 is an attestation report, not a certification. Live control status and report access: https://app.vanta.com/usetrident.dev/trust/cgsryaj8fquze0z74k9n0
- Integrations: AWS, Azure, Google Cloud, Kubernetes, GitHub, Bitbucket, Slack, Linear, Jira, and MCP.
- Website: https://tridentsecurity.io
- Demo: https://tridentsecurity.io/demo

## Definitions

Self-contained definitions for the concepts this site covers, written to be
quoted directly. Each links to the guide it is drawn from.

- **Cloud attack path analysis** — connecting the assets, identities, policies, network exposure, and data stores in a cloud account into the routes an attacker could walk between them, so remediation can be ranked by which change removes the most paths rather than by generic severity. See https://tridentsecurity.io/resources/cloud-attack-path-analysis
- **Continuous penetration testing** — an operating model that reruns authorized, evidence-producing tests when relevant systems or risks change, then verifies remediation by retesting. It is a lifecycle, not an unsupervised exploit loop. See https://tridentsecurity.io/resources/continuous-penetration-testing
- **Source code assisted pentest** — an authorized security test in which testers hold read access to the application source alongside a running instance. The code locates and explains candidate weaknesses; the running system is used to prove them. Reading code alone produces a code review, not a penetration test. See https://tridentsecurity.io/resources/source-code-assisted-pentest
- **AI penetration testing** — using large language models or autonomous agents to perform parts of a penetration test: enumerating attack surface, generating and adapting test cases, chaining steps toward an objective, and drafting findings. The distinction that matters is whether the tool validates exploitability against a running target or only predicts it. See https://tridentsecurity.io/resources/ai-penetration-testing
- **Toxic combination** — two or more conditions that are individually acceptable and jointly dangerous, such as a public workload paired with an over-broad instance role. See https://tridentsecurity.io/solutions/cloud

Note on scope: "AI pentesting" is used for two different things. This site
covers using AI to test conventional applications, APIs, and cloud
infrastructure. It does not cover penetration testing of AI models themselves.

## Product

- [AI Security for cloud and webapps](https://tridentsecurity.io/): Trident is an AI-powered security platform for continuous web and API penetration testing and cloud attack-path analysis across AWS, Azure, and GCP.
- [PR Scanning, Cloud Security & AI Pentesting](https://tridentsecurity.io/products): Three Trident products: PR scanning before merge, cloud attack paths across AWS, Azure, GCP and Kubernetes, and continuous web and API penetration testing.
- [One-Time Penetration Testing & Security Assessment](https://tridentsecurity.io/products/pentest): Trident’s security experts, accelerated by AI agents, take your web apps, APIs, source code and cloud apart once and walk you through what is actually exploitable. A report with reproducible evidence, a testing attestation and one retest within 30 days.
- [Cloud Security Attack Path Analysis](https://tridentsecurity.io/solutions/cloud): Map cloud assets, identities, and data stores, then connect exposure and access relationships into attack paths that help teams prioritize risk.
- [Web & API Penetration Testing](https://tridentsecurity.io/solutions/web-app): Test web applications and APIs for authentication, authorization, business-logic, and data-access risks, with findings organized for review.
- [Request a Demo — Trident Cloud Security & Pentesting](https://tridentsecurity.io/demo): Schedule a personalized Trident demo. See how teams map cloud attack paths, pentest web apps and APIs, and prioritize the risks that reach sensitive data.

## Docs

- [Trident Documentation](https://docs.usetrident.dev): Product documentation — quickstarts, integration guides and API reference. Machine-readable index at https://docs.usetrident.dev/llms.txt (full content at https://docs.usetrident.dev/llms-full.txt).

## Solutions

- [Security Solutions by Team & Industry](https://tridentsecurity.io/solutions): Cloud attack-path analysis and web and API penetration testing, mapped to the teams and regulated industries that rely on them.
- [Cloud Security & Pentesting for Startups](https://tridentsecurity.io/solutions/startups): Combine cloud attack-path analysis with web and API penetration testing to help startup teams identify and prioritize risks to sensitive data.
- [Enterprise Cloud Security & Pentesting](https://tridentsecurity.io/solutions/enterprise): Bring multi-cloud attack paths and web and API penetration testing into one view so teams can prioritize risks across business units.
- [Fintech API & Cloud Security Testing](https://tridentsecurity.io/solutions/fintech): Test payment applications and APIs for authorization and business-logic risks, then connect findings with cloud paths to sensitive financial data.
- [Banking Cloud Security & Pentesting](https://tridentsecurity.io/solutions/banks): Connect cloud risk, identity access, and web and API security findings to help banking teams prioritize exposure and organize evidence for review.
- [Healthcare Cloud & Application Security](https://tridentsecurity.io/solutions/healthcare): Map cloud paths to patient data and assess patient-facing web applications and APIs, helping teams prioritize security risks around PHI.
- [Cloud & App Security for HIPAA Programs](https://tridentsecurity.io/solutions/hipaa): Review cloud attack paths and web and API security findings around systems that handle PHI. Trident supports security work; it does not certify compliance.
- [Cloud Security & Pentesting for MSSPs](https://tridentsecurity.io/solutions/mssp): Explore cloud attack-path analysis and web and API penetration testing for client environments, with findings organized for review.

## Research and Blogs

- [Security Research and Vulnerability Write-ups](https://tridentsecurity.io/blog): Original vulnerability research from the team building Trident: how real exploits work, why they get past review, and the fix that closes each one.
- [Benchmarking Trident](https://tridentsecurity.io/blog/trident-web-pentest-benchmark): Claude Sonnet 4.6 confirmed a flaw in 36 of 44 runs on Trident, our pentest harness, and in 10 on Claude Code. DeepSeek V4 Flash on Trident confirmed one in 38 of 44, at an average of $1.29 per confirmed flaw. (published 2026-10-18)
- [A pristine file, a directory name, and arbitrary code](https://tridentsecurity.io/blog/pkgconf-env-shell-injection): pkgconf’s --env option wraps every value it prints in single quotes and then puts unescaped, attacker-influenced data between them. Three primitives escalate from a hostile .pc file, through an escaping routine written for the wrong shell context, to command execution driven by a directory name alone. (published 2026-09-12)
- [Freezing a six-figure crypto vault](https://tridentsecurity.io/blog/oracle-bounds-mismatch-vault-freeze): A few percent of disagreement between two contracts can freeze every deposit and withdrawal in a live crypto vault — and this one went unnoticed for six weeks. (published 2026-07-11)
- [A phone system, one request, and root](https://tridentsecurity.io/blog/cisco-cucm-rce): An unauthenticated SSRF in Cisco Unified CM’s WebDialer chains to a JSP webshell and root-level compromise of enterprise voice infrastructure. Cisco patched it on June 3; within weeks attackers were dropping webshells over Tor, and CISA gave federal agencies until June 28 to fix it. (published 2026-06-29)
- [Meta’s AI got tricked into resetting account passwords](https://tridentsecurity.io/blog/meta-ai-instagram-account-takeover): Attackers simply asked Instagram’s AI support assistant to send password-reset codes to an email they controlled, and it complied. (published 2026-06-01)

## Security guides

- [AI Penetration Testing: Methods & Limits](https://tridentsecurity.io/resources/ai-penetration-testing): Learn how pentesting with AI works, what it can automate, where human judgment remains essential, and how to evaluate evidence from an AI penetration test.
- [Cloud Attack Path Analysis Guide](https://tridentsecurity.io/resources/cloud-attack-path-analysis): Understand cloud attack path analysis, how identity and exposure become reachable risk, and how to prioritize choke points across AWS, Azure, and GCP.
- [Continuous Penetration Testing & Security](https://tridentsecurity.io/resources/continuous-penetration-testing): Learn how continuous security testing uses change-triggered penetration tests, reproducible evidence, remediation, and retesting instead of endless scanning.
- [Source Code Assisted Pentest: A Practical Guide](https://tridentsecurity.io/resources/source-code-assisted-pentest): What a source code assisted penetration test is, how it differs from black box and white box testing, what to give the testers, and how to judge the results.
- [How to Evaluate AI Pentesting Tools](https://tridentsecurity.io/resources/choosing-an-ai-pentesting-platform): A vendor-neutral framework for evaluating AI pentesting tools: exploit validation, evidence quality, scope control, and how to run a proof of concept.

## Optional

- [About Trident | AI Pentesting & Cloud Security](https://tridentsecurity.io/about): Learn how Trident, operated by Esprit Labs Inc., connects cloud attack-path analysis with web and API penetration testing for security teams.
- [Privacy Policy](https://tridentsecurity.io/privacy): How Esprit Labs Inc. collects, uses, and protects information when you use the Trident services.
- [Terms of Service](https://tridentsecurity.io/terms): Standard terms of service governing access to and use of the Trident services provided by Esprit Labs Inc.