Arcjet Launches Runtime Security for AI Agents

Arcjet announces in a press release an agent runtime security product for monitoring activity, enforcing policies, and retaining audit evidence. It connects actions across sessions to show each agent workflow as a sequence rather than as separate events.

The product can ingest activity through existing OpenTelemetry tools without application code changes. It can also retrieve activity from the Claude Compliance API and record prompts, tool parameters, session metadata, identity, and security decisions.

Security teams can apply controls before and after calls to language models, tools, databases, and APIs. Policies powered by Rego and Open Policy Agent can address prompt injection, sensitive data exposure, rate limits, quotas, approved recipients, refund values, and trusted URLs.

Arcjet supports the Claude Agents SDK, Claude Managed Agents, OpenAI Agents SDK, LangChain, LangFuse, Strands, Mastra, and Microsoft Agent Framework. Correlated traces preserve actions, inputs, policy evaluations, and decisions for security reviews and compliance audits.