Arcjet Agent Runtime Security

Arcjet Agent Runtime Security is an in-code runtime layer that discovers, controls, and audits AI agent behavior in production, checking each LLM, tool, and database call against deterministic policy, with native support for the Claude Agents SDK, OpenAI Agents SDK, and LangChain.

Arcjet Agent Runtime Security, from developer security company Arcjet, addresses a gap that opens once AI agents move from demos into production: agents can write to databases, answer support tickets, and process refunds, and those workflows often start from an email or a code commit and then move across several systems with nobody checking each step. Arcjet gives teams observability, enforcement, and audit across those agent workflows so they can see which agents are active, control what each one can do, and reconstruct what happened afterward. The product works in code, wrapping the calls an agent makes to models, tools, and databases so each action can be evaluated against policy before it runs. Teams can set deterministic rules to detect prompt injection, redact personal information before it leaks, and apply bot detection and rate limits. Native framework support covers the Claude Agents SDK, the OpenAI Agents SDK, and LangChain, and platform teams can route agent activity through the OpenTelemetry tooling they already run. Claude customers also have a path through the Anthropic Compliance API. Every action produces audit evidence that security and compliance teams can review. Arcjet was founded in 2023 by David Mytton and is based in San Francisco. It builds developer-focused security that integrates directly into the application codebase rather than sitting only at the network edge. Agent Runtime Security extends that in-code approach from web apps to autonomous agents, and it is available now with a free trial.