A San Francisco cybersecurity startup has launched an artificial intelligence agent that penetrates computer systems, identifies vulnerabilities, and then autonomously generates code to repair what it finds. Antigen, backed by Lightspeed Venture Partners and Y Combinator, raised nearly $4 million at a $30 million valuation in a seed round that also drew investment from DST Global and Susa Ventures, according to co-founder Saad Jamal in an August video posted to LinkedIn, though the funding details have not been independently verified.
The product, called tCell, enters a market that has become unexpectedly crowded over the past year. At least seven established security vendors and newer entrants shipped similar autonomous penetration testing tools in recent months, but Antigen is positioning itself as offering something tighter: a continuous loop that runs from attack simulation through remediation and back to verification.
tCell works by mapping an organization's digital attack surface and then chaining individual security weaknesses into what the company describes as exploitable paths capable of reaching sensitive data. Unlike traditional vulnerability scanners that produce static reports, each finding from tCell includes a working exploit that security teams can replay to understand exactly how an attacker might gain access.
"We build an AI agent that hacks into you, finds security vulnerabilities and fixes them," Jamal explained in the video.
After identifying an exploitable route, the agent writes a proposed code fix and opens a pull request in GitHub or another code repository. The patch goes to human engineers for review. Nothing merges without approval. Once deployed, tCell reruns its exploit to confirm the vulnerability has closed, and if that retest reveals lingering exposure, the agent reopens the issue with a revised patch.
The agent can navigate some of the tedious obstacles that typically require human intervention during security testing. According to Antigen's documentation, tCell signs up for accounts, authenticates using Google single sign-on, and solves CAPTCHAs to reach protected sections of applications. The company says agents run only against approved targets and operate with what it calls "skills and guardrails," though public documentation does not spell out specific controls or kill switches.
From Google and Tesla to Hacking Agents
Abdullah Nauman and Jamal founded Antigen after working at Google and Tesla. Nauman spent time on product at Google Ads and machine learning at Google Search, while Jamal built infrastructure for Tesla's Autopilot system and earlier trained reasoning models alongside Ashish Vaswani, according to the company's Y Combinator profile. The startup went through a recent Y Combinator batch and emerged from stealth around June, Jamal said.
Corporate filings updated in early August list the legal entity as Claybird Inc., operating as Antigen. Y Combinator's jobs board lists the team size as five, though the company's LinkedIn page indicates 11 to 50 employees. The jobs board shows five open positions.
Deployment Options and Enterprise Integrations
Antigen offers both managed software-as-a-service and on-premises deployment through what it calls "Dedicated" workers that run inside customer networks on Amazon Web Services, Microsoft Azure, Google Cloud, or Kubernetes environments. The platform connects with GitHub, Linear, Jira, and ServiceNow for ticketing, and sends audit logs to Splunk, Datadog, Elastic, Microsoft Sentinel, and Panther, according to the company's enterprise documentation.
A page aimed at government customers claims the product supports reporting aligned with PCI DSS requirement 11.4, SOC 2, NIST frameworks, and FISMA authorization packages, and lists procurement availability through reseller Pursuit. The company has not disclosed formal third-party certifications.
An Increasingly Busy Category
The autonomous pentesting market has seen a flurry of launches. Snyk introduced Evo Continuous Offensive Security in early August. Synack announced general availability of Sara AI Pentesting in May, and NetSPI released AI-powered continuous pentesting around the same time; the two companies later merged in September. Pentera unveiled AI-native web application pentesting in late July, YesWeHack launched Agentic Pentest in late June, and smaller players including SpartanX and Assail went live earlier in the year.
Several other startups, among them Strix, Pensar, and Adversarial.sh, also advertise agents capable of exploiting vulnerabilities and proposing fixes, based on their publicly available websites.
Antigen declined to name customers when contacted. The company's website includes a section labeled "Trusted by teams at" that contains no logos, and no public case studies were available as of early October. That absence is perhaps not unusual for a company this young, though it leaves little external validation of how the technology performs in production environments.
Jamal indicated in his August video that the seed round came together with investors joining "piecemeal" over time rather than in a single close. The company is currently hiring for security researcher roles, with salary and equity compensation bands posted on Y Combinator's jobs board.
