About Kontext
Kontext is a runtime security platform for AI agents that sits between agents and the tools and systems they use. It evaluates each agent action in real time against security policies, considering identity, requested action, target resource and task. Teams can start in observe mode and later block unauthorised actions with an auditable record.
What is Kontext?
What is Kontext?
Kontext is an AI security company building a runtime control platform for AI agents. As agents move beyond chat into workplace environments where they write code, access files and act with company credentials, organisations need a way to control what those agents are allowed to do while they run.
Kontext's software sits between AI agents and the tools and systems they interact with. It evaluates agent activity in real time against security policies and risk signals, taking into account the agent's identity, the requested action, the target resource and the assigned task. Its website describes checking every Claude Code, Codex and Cowork tool call against policy before it runs.
Teams can first deploy Kontext in observe mode to see how agents behave, spot potentially risky activity and understand how policies would apply without interrupting work. When enforcement is switched on, the platform can block unauthorised actions before they execute and keep an auditable record of each decision.
Kontext was founded by Jens Ernstberger and Michel Osswald, whose backgrounds include secure computing, applied cryptography and AI systems. In September 2026 it raised USD 4 million led by 42CAP, with backing from a16z CSX and HTGF, to expand engineering and develop its runtime enforcement platform.
Pros
- Evaluates agent actions before they execute
- Considers identity, action, resource and task context together
- Observe mode allows testing policies without disrupting work
- Keeps an auditable record of each decision
Cons
- Focused on organisations already running AI agents
- Adds a control layer that teams need to configure and maintain
- Early-stage company with limited public customer information
How to Use Kontext
- 1Get in touch
Request access through the Kontext website.
- 2Deploy in observe mode
Connect Kontext between your agents and tools to watch how agents behave.
- 3Write policies
Define which actions agents may take based on identity, resource and task.
- 4Enable enforcement
Switch on blocking so unauthorised actions are stopped before they run, and review the audit log.