Vulnerability Management
Code Security
Cloud Security
Threat Hunting

The agent harness for cybersecurity

Automate workflows, investigate vulnerabilities, track down owners, and remediate fast — at scale, with zero engineering friction.

Book a demo

Adversaries are getting faster

AI turns a CVE into a working exploit in hours, and anyone can use it. Are you ready for the vulns of the future?

AnthropicSep 29, 2026

An open model wrote a Chrome exploit in 8 hours

An open-weight model turned a public Chrome patch into a working exploit in 8 hours, for $20.40.

Anthropic research

JFrog ArtifactoryAug 28, 2026

From fix to 400K attacks a day in 5 days

One request gave attackers admin on the server that ships your code. Two weeks on, half hadn't patched.

Fastly

2026 zero-days in production stacksPaused

Next.jscritical fix withheldSep 30
AF_UNIX escapeCVE-2026-80521Sep 22
Next.js next/ogCVE-2026-94545Sep 22
Linux ebtablesCVE-2026-53266Sep 18
Windows UpdateCVE-2026-81963Sep 8
Windows ALPCCVE-2026-85880Sep 8
Adobe CommerceCVE-2026-75650Sep 7
Next.js on WindowsCVE-2026-75604Aug 25
libheif AVIFvia next/imageAug 25
MLflowCVE-2026-64849Aug 19
Windows AFDCVE-2026-68820Aug 11
LangflowCVE-2026-9198Aug 4
Apache TomcatCVE-2026-34486Aug 4
Fastjson 1.xCVE-2026-16723Jul 21
wp2shellCVE-2026-63030Jul 20
Windows ADFSCVE-2026-56155Jul 14

AnthropicSep 29, 2026
An open model wrote a Chrome exploit in 8 hours
An open-weight model turned a public Chrome patch into a working exploit in 8 hours, for $20.40.
Anthropic research
2026 zero-days in production stacksPaused
- Next.jscritical fix withheldSep 30
- AF_UNIX escapeCVE-2026-80521Sep 22
- Next.js next/ogCVE-2026-94545Sep 22
- Linux ebtablesCVE-2026-53266Sep 18
- Windows UpdateCVE-2026-81963Sep 8
- Windows ALPCCVE-2026-85880Sep 8
- Adobe CommerceCVE-2026-75650Sep 7
- Next.js on WindowsCVE-2026-75604Aug 25
- libheif AVIFvia next/imageAug 25
- MLflowCVE-2026-64849Aug 19
- Windows AFDCVE-2026-68820Aug 11
- LangflowCVE-2026-9198Aug 4
- Apache TomcatCVE-2026-34486Aug 4
- Fastjson 1.xCVE-2026-16723Jul 21
- wp2shellCVE-2026-63030Jul 20
- Windows ADFSCVE-2026-56155Jul 14
JFrog ArtifactoryAug 28, 2026
From fix to 400K attacks a day in 5 days
One request gave attackers admin on the server that ships your code. Two weeks on, half hadn't patched.
Fastly

Solution

Solve the hardest problems in cybersecurity using frontier intelligence

The Agent Harness gives frontier models the subagents and context of your stack to do real security work.

SubagentsContext

Agent Harness

Subagents

Agent Harness

Context

Partner
Built for you
- A dedicated Rival engineer tailors agents to your environment
Agent Store
Ready-made agents
- Agents you can deploy today
- Tailored on installation
Grow
Build your own
- Easily automate any complex work
- Your agents stay yours

We’re on

Meet the @rival harness

Ask Rival anything, or pick a built-in agent to start a workflow.

What should we secure today?

Built-in agents

Recent Work by

research

Mythos 'Discovered' a CVE Already in Its Training Data — and That's Still Worrying

Anthropic made headlines claiming Claude Mythos achieved the "first remote kernel exploit discovered and exploited by an AI." We went looking for how — and found a 20-year-old bug hiding in plain sight.

research

Taxi: Debugging Agent Trajectories at Scale

Behavioral bugs in agents fail silently and repeatedly, quietly burning tokens, time, and quality. Taxi is how we find them — clustering agent trajectories into a taxonomy we can query, browse, and act on.

research

ClosedCaption: Finding Interpretable Clusters with LLMs

We discuss use cases for LLM-cluster interpretability in agent analysis, experimental designs to test these pipelines, and how we use these techniques to analyze our own agents.

Show all posts

The agent harness for cybersecurity

Adversaries are getting faster

An open model wrote a Chrome exploit in 8 hours

From fix to 400K attacks a day in 5 days

An open model wrote a Chrome exploit in 8 hours

From fix to 400K attacks a day in 5 days

Solve the hardest problems in cybersecurity using frontier intelligence

Built for you

Ready-made agents

Build your own

Meet the @rival harness

What should we secure today?

Recent Work by

Mythos 'Discovered' a CVE Already in Its Training Data — and That's Still Worrying

Taxi: Debugging Agent Trajectories at Scale

ClosedCaption: Finding Interpretable Clusters with LLMs

Every change needs a beginning.