<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw" -->

---
title: Google Mantis: An Agentic Vulnerability Scanning Harness...
description: Google has open-sourced Mantis, an agentic AI framework built to automate vulnerability discovery, validation, reproduction, and patching. It aims to fix the...
canonical: https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Google Mantis: An Agentic Vulnerability Scanning Harness for Reducing False Positives | daily.dev
og:description: Google has open-sourced Mantis, an agentic AI framework built to automate vulnerability discovery, validation, reproduction, and patching. It aims to fix the...
og:url: https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw
og:image: https://api.daily.dev/og/posts/WRVjSQogw.png
og:image:alt: Google Mantis: An Agentic Vulnerability Scanning Harness for Reducing False Positives
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Google Mantis: An Agentic Vulnerability Scanning Harness for Reducing False Positives

**[InfoQ](https://daily.dev/sources/infoq)** · 3 min read · 1 upvotes · 0 comments

## Summary

Google has open-sourced Mantis, an agentic AI framework built to automate vulnerability discovery, validation, reproduction, and patching. It aims to fix the weak true-positive rates (under 7%) and hallucinated bugs common in conventional AI code scanning by combining critic and reviewer agents with sandboxed reproduction for grounding. Mantis analyzes repository history, architecture, and threat models, summarizing files into a hierarchical context tree that cuts token usage by 85%. The framework includes over 15 modular tools (mantis-summarize, mantis-review, mantis-critic, mantis-reproduce, mantis-patch, and more) that can run sequentially or in parallel, communicating via shared on-disk state. It supports mixing model classes per stage, using lighter models for simple classification and heavier models for reproduction and patch generation. Mantis is available on GitHub with an accompanying agent reference guide.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://www.infoq.com/news/2026/09/google-mantis-vulnerability-scan>

## Questions this post answers

### What is Google Mantis and how does it reduce false positives in AI vulnerability scanning?

Mantis is an open-source AI-agent framework from Google that automates finding, validating, reproducing, and fixing software vulnerabilities. It combines critic and reviewer agents with sandboxed reproduction of vulnerabilities to ground findings in evidence rather than relying solely on an LLM's judgment, addressing conventional AI scanners' true-positive rates often below 7%.

_Teams evaluating AI-driven vulnerability scanners can track tools like Mantis as they evolve on daily.dev._

### How does Mantis reduce token usage when analyzing large codebases?

Mantis summarizes analyzed files into a hierarchical tree containing directory- and repository-level context instead of scanning files with brute force. This structural summarization reduces token usage by 85% while retaining the important context needed for the agents to reason about vulnerabilities.

_Developers optimizing LLM pipeline costs can follow techniques like this on daily.dev._

### Which AI model sizes should be used for different stages of an agentic vulnerability scanning pipeline?

Lighter 'flash' or 'lite' model variants suit tasks without deep logic requirements, such as rapid classification in mantis-researcher or clustering similar patterns in mantis-dedupe. Heavier, more advanced models are better for tasks needing deep contextual understanding, such as mantis-reproduce, which writes functional crash reproducers, and mantis-patch, which generates side-effect-free code fixes.

_Engineers designing multi-model agent pipelines can compare approaches like this on daily.dev._

## Similar posts on daily.dev

- [Using AI agents to secure Google infrastructure](https://daily.dev/posts/using-ai-agents-to-secure-google-infrastructure-n2sosodh7) · Google Cloud · 0 upvotes · 0 comments
- [Build your own vulnerability harness](https://daily.dev/posts/build-your-own-vulnerability-harness-qm0qjudbn) · Cloudflare · 4 upvotes · 0 comments

---

Tags: [#security](https://daily.dev/tags/security), [#google](https://daily.dev/tags/google), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#appsec](https://daily.dev/tags/appsec)

[View this post on daily.dev](https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Google Mantis: An Agentic Vulnerability Scanning Harness for Reducing False Positives","url":"https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw"},"datePublished":"2026-09-06T12:02:54.134Z","dateModified":"2026-09-14T08:35:45.237Z","description":"Google has open-sourced Mantis, an agentic AI framework built to automate vulnerability discovery, validation, reproduction, and patching. It aims to fix the...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/7a365d7e68a163a3e9c1d2c3104d0ced?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/7a365d7e68a163a3e9c1d2c3104d0ced?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"InfoQ","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"InfoQ","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/afc3bced3e1e4b188dd9127017a60e0c","url":"https://daily.dev/sources/infoq"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,google,llm,ai-agents,appsec","timeRequired":"PT3M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"InfoQ","item":"https://daily.dev/sources/infoq"},{"@type":"ListItem","position":3,"name":"Google Mantis: An Agentic Vulnerability Scanning Harness for Reducing False Positives"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/google-mantis-an-agentic-vulnerability-scanning-harness-for-reducing-false-positives-wrvjsqogw#faq","mainEntity":[{"@type":"Question","name":"What is Google Mantis and how does it reduce false positives in AI vulnerability scanning?","acceptedAnswer":{"@type":"Answer","text":"Mantis is an open-source AI-agent framework from Google that automates finding, validating, reproducing, and fixing software vulnerabilities. It combines critic and reviewer agents with sandboxed reproduction of vulnerabilities to ground findings in evidence rather than relying solely on an LLM's judgment, addressing conventional AI scanners' true-positive rates often below 7%. Teams evaluating AI-driven vulnerability scanners can track tools like Mantis as they evolve on daily.dev."}},{"@type":"Question","name":"How does Mantis reduce token usage when analyzing large codebases?","acceptedAnswer":{"@type":"Answer","text":"Mantis summarizes analyzed files into a hierarchical tree containing directory- and repository-level context instead of scanning files with brute force. This structural summarization reduces token usage by 85% while retaining the important context needed for the agents to reason about vulnerabilities. Developers optimizing LLM pipeline costs can follow techniques like this on daily.dev."}},{"@type":"Question","name":"Which AI model sizes should be used for different stages of an agentic vulnerability scanning pipeline?","acceptedAnswer":{"@type":"Answer","text":"Lighter 'flash' or 'lite' model variants suit tasks without deep logic requirements, such as rapid classification in mantis-researcher or clustering similar patterns in mantis-dedupe. Heavier, more advanced models are better for tasks needing deep contextual understanding, such as mantis-reproduce, which writes functional crash reproducers, and mantis-patch, which generates side-effect-free code fixes. Engineers designing multi-model agent pipelines can compare approaches like this on daily.dev."}}]}
```

