Superagent blog · Mar 2026
Published a benchmark showing frontier models missed 57.3% of threats already identified by brin across 485 real artifacts.
Company research
Public company, workplace, funding, and market signals
Updated Jul 30, 2026
Overview
Superagent is a Y Combinator W24 AI security platform that continuously finds, patches, and discloses vulnerabilities in code and AI agents, with fixes shipped as GitHub pull requests and a human-in-the-loop approval flow.
Primary product
AI security platform for continuous find-and-fix vulnerability research on code and agents
Founded
2024
Headquarters
San Francisco, California, United States
Team size
1-10 employees
Work style
Remote
Industry
Technology, Information and Internet
Sub-industry
AI security / agent security
Offices
San Francisco, California, United States
Dover, Delaware, United States (legal/registered address)
Sweden
India
Business model
open-source freemium
B2B SaaS
enterprise/custom pricing
usage-based API
Funding
People & investors
Work & culture
Security-first, research-heavy, open-source, and human-in-the-loop, with a distributed team and a bias toward shipping fixes quickly as PRs.
Work style
Remote
Product & market
Pricing
Free for public repositories; custom quote for private repositories and teams; usage-based API pricing for safety methods.
Differentiators
- —Ships remediation as GitHub pull requests
- —Continuous find-and-fix security workflow
- —Human-in-the-loop approval before fixes merge
- —Free for public repositories
- —Security coverage for both code and AI agents
- —Integrates into GitHub/CI workflows without extra tooling
Technology
GitHub Apps
Dropbox Sign
Stripe
Vercel
Supabase
PostHog
Loops.so
AWS
Customers
dotenvx
Capchase
Competitors
Semgrep
Snyk
GitHub Advanced Security
CodeRabbit
Lakera
Scale & contact
Estimated monthly visits
40.8K
Latest news
Superagent blog · Feb 2026
Argues that the Cline incidents show why treating untrusted content as instructions is a broken security model for AI agents and software supply chains.
Superagent blog · Feb 2026
Introduced brin.sh, a realtime context trust scoring service for agents that scores packages, web pages, repos, and MCP servers before agents trust them.
Superagent blog · Jan 2026
Released three guard models for runtime detection of prompt injections and jailbreaks, with local/VPC deployment and structured pass/block outputs.
Superagent blog · Sep 2025
Announced Superagent as a runtime defense platform for AI agents, highlighting protection against prompt injections, malicious tool calls, and data leaks.
Jobs and careers at Superagent
0 jobs at Superagent
No open positions
Check back later for new openings