Arcjet Labs Inc. has introduced a new runtime security tool designed to monitor and manage artificial intelligence agents operating in production environments. The platform tracks active agents and validates their actions against predefined security policies before execution.
The system integrates with existing tools like OpenTelemetry and the Anthropic Compliance API, allowing teams to view agent activity in a centralized inventory. It supports various frameworks, including the Claude Agent SDK, OpenAI Agents SDK, LangChain, Mastra, and Microsoft Agent Framework.
Security teams can implement rules to detect prompt injections, redact sensitive data, and set rate limits. Policies are based on Open Policy Agent and can be updated without redeploying applications, enabling real-time adjustments to agent behavior.
The company also updated its Python and Go software development kits to version 1.0. Currently, Arcjet software is utilized by over 500 production applications. The startup was founded three years ago by David Mytton.
Arcjet secured $8.3 million in a Series A funding round led by Plural last October. Other investors include Andreessen Horowitz, Seedcamp, and former Twilio CTO Ott Kaukver.
