Skip to main content
Contact on WhatsApp

Meet 0: The Open-Source AI Security Harness That Lets AI Agents Do Real Security Work

Meet 0: The Open-Source AI Security Harness That Lets AI Agents Do Real Security Work

0 is an open-source AI security harness from 0sec-labs, backed by Y Combinator and the Swiss Applied AI and Cybersecurity Research Lab. It gives you a browser console where you can investigate vulnerabilities, review findings, and work on fixes, using your own model provider. Local use needs no account: install it on Apple Silicon macOS or x64 and ARM64 Linux, run the command 0, and the console opens in your browser. One harness covers code analysis, penetration testing, vulnerability research, reverse engineering, verification, and fixing, with a CLI and MCP access so coding agents can run the same workflows you run by hand.

Inside the console, findings carry source evidence and sit in a queue or on a board, so a team can triage what matters instead of drowning in scanner output. Built-in workflow templates cover assessment, verification, fixes, and research, and every run keeps its results, while Learning keeps local activity and source-grounded notes you can reuse. For automation, the CLI exposes the same workflow runner: list templates, run a security-research workflow against a repository, list runs, or chat with a prompt. Agents connect through MCP, with local MCP running under an explicitly selected host-local execution profile. The 0verse add-on brings local binary analysis through Ghidra, with nine MCP tools for scanning, triage, preflight, bug classes, findings, proofs of concept, reports, memory recall, and capability discovery. A hosted version, 0cloud, is on a waitlist, but local use needs no cloud account.

Why this matters for AI security engineers and red teamers: security work is moving from chatbots that describe vulnerabilities to agents that run real workflows, keep evidence, and produce fixes. That power makes tool permissions and scope the real safety boundary, which is why 0 ships a scope plugin that enforces target boundaries. Two honest limits: it is a research preview, and it must only be used on authorized systems, with findings and fixes reviewed before action. It is open source under MIT or Apache-2.0, and it is already drawing attention on GitHub. If you want to build this kind of skill, from LLM agent security to hands-on red teaming, explore the trainings at https://cyberlegends.org and follow Cyber Legends for more tool breakdowns. Similar repositories worth watching, all open source and highly starred on GitHub (star counts as checked on 3 November 2026): 1. Strix - https://github.com/usestrix/strix - Open-source AI penetration testing tool to find and fix your app's vulnerabilities. 73,426 stars, Apache-2.0. 2. promptfoo - https://github.com/promptfoo/promptfoo - Test your prompts, agents, and RAGs. Red teaming, pentesting, and vulnerability scanning for AI. 26,760 stars, MIT. 3. DeepEval - https://github.com/confident-ai/deepeval - The LLM Evaluation Framework. 19,258 stars, Apache-2.0. 4. PentestGPT - https://github.com/GreyDGL/PentestGPT - Automated penetration testing agentic framework powered by large language models. 16,248 stars, MIT. 5. garak - https://github.com/NVIDIA/garak - The LLM vulnerability scanner. 9,839 stars, Apache-2.0. 6. NeMo Guardrails - https://github.com/NVIDIA-NeMo/Guardrails - Open-source toolkit for adding programmable guardrails to LLM-based conversational systems. 7,419 stars. 7. PyRIT - https://github.com/microsoft/PyRIT - Open-source AI red teaming toolkit from Microsoft for testing generative AI systems. 4,731 stars, MIT. 8. Inspect AI - https://github.com/UKGovernmentBEIS/inspect_ai - A framework for large language model evaluations. 3,158 stars, MIT. 9. Vulnhuntr - https://github.com/protectai/vulnhuntr - Zero shot vulnerability discovery using LLMs. 2,815 stars, AGPL-3.0. 10. HarmBench - https://github.com/centerforaisafety/HarmBench - A standardized evaluation framework for automated red teaming and robust refusal. 1,083 stars, MIT.

Share Article

Related Insights

The Great DM Decryption: 5 Impactful Takeaways fr…
Mar 14, 2026
The Obsolete Roadmap: 6 Counter-Intuitive Truths …
Mar 16, 2026
5 Surprising Realities of the 2026 Cyber Landscap…
Mar 16, 2026
Link copied to clipboard!
Footer - Cyber Legends