by Nebula Security
Autonomous AI agent that finds and patches real vulnerabilities in Chrome, Linux, and code.
Vega fits security teams shipping large C/C++ codebases, think browser engines, kernels, and other systems-level infrastructure, who would rather run continuous automated scans than periodic manual audits. One of its Linux kernel finds earned a $10,500 kernelCTF bounty, real evidence its exploits are exploitable, not academic. It replaces a first pass of manual research; a human reviews findings before disclosure.
Fully autonomous exploit chain to full V8 sandbox escape and RCE on a real Chrome CVE, without any human-written exploit code.
A real, publicly verifiable CVE track record (3 named Chrome/Linux CVEs, 730+ bugs) is hard for a new entrant to fake, but the underlying model is undisclosed proprietary tooling with no open benchmark, so the moat rests on reputation more than a defensible technical edge.
Main trade-off
No published flat pricing and a very narrow domain focus (C/C++ systems code) limit it to a small buyer pool of teams shipping browsers, kernels, or similarly low-level software.
Vega Scan is Nebula's self-serve tier: usage-based pricing tied to a repository's size and complexity, running automated vulnerability hunting, root cause analysis, patch generation, and PR review with no manual sign-off required to start. Premium Audit drops the self-serve model for a custom-quote engagement, pairing a senior Nebula security engineer with Vega to review architecture, model threats, and audit on-premise deployments. Neither tier lists a public price, and there is no free plan.
| System | Type |
|---|---|
| GitHub | Native |
| Claude Code | Plugin |
| Cursor | Plugin |
Vega is an AI security agent from Nebula Security that autonomously hunts, root-causes, and patches vulnerabilities in C/C++ systems code. Its public research catalog credits it with 730+ documented bugs, including three named Chrome V8 CVEs in 2026. It ships as a GitHub App, a CLI, and an MCP server for Claude Code and Cursor.
Autonomous vulnerability discovery
Vega has found and reported real bugs across major codebases, including the Linux kernel and Chrome V8, without a human writing the exploit chain.
Full sandbox escape exploits
On CVE-2026-6307, Nebula reports Vega achieved arbitrary memory read/write and a complete V8 heap sandbox escape to RCE with a 100 percent success rate on its own, no additional bug chaining needed.
Root cause analysis and patch generation
Vega Scan does not just flag a bug; it traces the root cause and generates a candidate patch alongside a proof of concept.
GitHub App pull request review
Vega reviews incoming pull requests automatically as a GitHub App, catching vulnerable code before it merges.
MCP integration for agentic coding
An MCP server exposes Vega inside Claude Code, Cursor, and other agent-driven IDEs so security scanning runs in the same loop as code generation.
90+30 day responsible disclosure
Every vulnerability Vega finds follows Nebula's standard 90-day vendor notice plus 30-day public disclosure window, tracked in a public research catalog of confirmed CVEs.
| Agent | Where it runs | Price | Maturity | Checked |
|---|---|---|---|---|
| Cloud · API | Usage-based | Available | 26 Aug | |
| -- | -- | -- | -- | |
| Cloud · API | Free plan | Available | 25 Oct | |
| Cloud | Quote | Alpha | 26 Aug |
Vega has no published flat price. Vega Scan is metered by usage, scaled to a target repository's size and complexity, and covers fully automated hunting, root cause analysis, patch drafting, and pull request review once you sign up. Premium Audit is the custom-quote alternative: a senior Nebula security engineer works directly alongside Vega's analysis on deeper architecture and threat-modeling work. Budgeting for either tier means talking to Nebula first, since nothing is listed publicly.
No, Vega has no free plan: every path into the product, Vega Scan or Premium Audit, requires a paid engagement. There is no trial run or capped usage tier to test it against your own codebase before signing up. Teams wanting to evaluate it first would need to start a conversation with Nebula's sales team directly.
XBOW and RunSybil are Vega's closest competitors, though both take a broader approach than Vega's narrow systems focus. XBOW does black-box web application penetration testing at scale and reached number one on HackerOne's leaderboard in 2025 with over 1,060 submitted vulnerabilities. RunSybil raised $40M to build Sybil, a multi-agent orchestrator that runs full black-box pentests, recon through exploit chaining, without human intervention. Vega stays narrower by design: browser engines, kernels, and other low-level C/C++ systems code.
XBOW and Vega solve different problems. XBOW hunts broadly across web applications in a black-box style, with a HackerOne track record proving it at scale. Vega goes narrower and deeper, specializing in C/C++ systems code such as browser engines and kernels, where it has produced named CVEs instead of bug bounty submissions. A team building a typical web app gets more mileage from XBOW; one shipping browser or kernel code gets more from Vega's domain-tuned pipeline.
Start by signing up for Vega Scan on nebusec.ai and connecting the target repository; pricing gets scoped from there and automated hunting begins right away. Teams that want pull requests reviewed automatically should install the Vega GitHub App, and teams running CI/CD can add the CLI to their pipeline instead. Anyone already working inside Claude Code or Cursor can connect Vega's MCP server so scanning runs in the same agentic coding session. Premium Audit starts differently, with a consultation to scope the engagement before Nebula quotes a price.
Vega is the flagship AI agent built by Nebula Security, a San Francisco based, Y Combinator backed security research lab founded in 2022. Unlike a general-purpose chatbot bolted onto static analysis tools, Vega bakes security domain knowledge directly into its agentic workflow: it reads source code, forms hypotheses about exploitable flaws, writes and runs proof-of-concept exploits, and validates its own findings before a human ever sees them.
What makes Vega an agent rather than a scanner is that it chains multi-step reasoning end to end, moving from recon over a codebase through root cause analysis, patch generation, and proof-of-concept validation, all before opening a pull request. Nebula's public research catalog credits Vega with three named 2026 CVEs, CVE-2026-6307 and CVE-2026-5865 in Chrome's V8 engine among them, plus a Linux kernel netfilter bug, CVE-2026-23274, that won a $10,500 kernelCTF bounty. Nebula claims the narrow, domain-tuned agentic loop behind these finds outperforms equivalent general-purpose GPT and Claude based security setups on Chromium and Linux kernel codebases specifically.
Vega is built for security teams and engineering orgs that ship large, security-critical C/C++ and systems codebases, think browser engines, kernels, and other low-level infrastructure, and want continuous automated vulnerability hunting instead of periodic manual audits. It is not a general chatbot assistant and is not aimed at casual developers; it is aimed at teams already running bug bounty programs or internal security research who want an AI agent doing the first pass of exploit discovery.
Autonomy, guardrails, integrations and pricing on this page are read from Nebula Security's own documentation. The verdict is Hokai's reading of them, not the vendor's claim.