AI agents at work can write code, open files and use passwords. Kontext, a startup from Munich, wants to check each of those actions before it happens. It has raised $4 million in a round led by 42CAP, with a16z CSX and HTGF also investing.
Kontext was founded by Jens Ernstberger and Michel Osswald. The company says the money will go into hiring engineers and building out its platform.
Why it matters
Security tools today mostly check who is logged in and which apps they may use. An agent can pass all of those checks and still do something nobody wanted. "An AI agent can be properly authenticated, use an approved tool, and still take an action no one authorized," Ernstberger said in the announcement.
Julian von Fischer of 42CAP made the same point from the investor side: the identity and access tools of the last twenty years "assume a human is on the other end, clicking one thing at a time."
What the product does
Kontext sits between an agent and the systems it touches. Each requested action is weighed against policy: which agent is asking, what it wants to do, on which resource and for which task. Companies can first run it in a watch-only mode and later let it block actions that break the rules, with a record of each decision.
The startup says it already checks actions from coding agents such as Claude Code and Codex.
