Straiker Raises $64 Million to Secure Enterprise AI Agents

Straiker, a Sunnyvale company that calls itself the agentic security company, said on June 29, 2026 that it raised a $64 million Series A. That brings total funding to $85 million. The money has one job: protect the AI agents that companies now hand real access to real systems.

The round was led by Marathon Management Partners, with Citi Ventures, Illuminate Financial, and Workday Ventures joining, plus existing backers Bain Capital Ventures and Lightspeed. The founders are not newcomers. CEO Ankur Shah ran Prisma Cloud at Palo Alto Networks, and CTO Sreenath Kurupati led AI and security research at Akamai.

Why a security round is suddenly big news

For two years the AI story was about capability: what agents can do. This round is about the other half of the story, what happens when they do it inside your bank, your clinic, or your call center. IDC projects more than one billion AI agents deployed across enterprises by 2029, roughly forty times the 2025 number. An agent is not a chatbot. It holds credentials, calls tools, moves data, and increasingly moves money.

Straiker put numbers on the risk. In its own testing, 91 percent of attacks on productivity agents ended in silent data exfiltration, and 36 percent of successful attacks on coding agents led to remote code execution. The company also flagged that 28.6 percent of cataloged MCP tools, the connectors agents use to reach other software, are dangerous by design. "AI agents are getting the keys to real systems and real data, which makes securing them one of the defining challenges of this era," said Brad Jones, a chief information security officer quoted in the announcement.

What the product actually does

Straiker sells three pieces. Discover AI finds every agent already running inside a company and who owns it. Ascend AI runs adversarial red teaming before an agent ships, throwing prompt injection, tool poisoning, and memory tampering at it to see what breaks. Defend AI sits at runtime and blocks the same attacks in production. It covers coding agents like Claude Code and Cursor, productivity agents like ChatGPT and Copilot, and custom agents built on LangChain or AWS Bedrock. The company reports run rate revenue grew fifteen times in under a year, and CEO Ankur Shah told investors that "demand is outpacing anything we forecast."

What this means for banks, clinics, and call centers

At PATech we build voice AI agents that answer real phone lines, book appointments, and pass leads into a business. Our agents touch calendars, customer records, and payment flows. So this news is not abstract for us, it is the exact problem we design around every day.

Three practical lessons carry over to any team putting an agent into production. First, know your agents. You cannot secure what you have not inventoried, and the shadow agents that staff spin up on their own are the ones that leak. Second, red team before you launch, not after. An agent that passes a friendly demo can still hand data to a hostile prompt, so test it the way an attacker would. Third, keep a guardrail at runtime. Static rules written last quarter will not catch a live prompt injection, so the check has to run on every request, in real time.

There is a compliance angle too. A bank agent that can move money, the kind of agentic payment flow we covered recently, deserves the same scrutiny a junior employee with system access would get. A clinic agent that reads patient records lives under HIPAA whether or not anyone remembered to check.

The takeaway

Straiker did not raise $64 million because agents are a fad. It raised it because agents are becoming coworkers with logins, and every coworker with a login is a security surface. The teams that win with agents will treat security as part of deployment, not a fix bolted on after the first incident. Sources: PR Newswire, Axios, TAMradar.