ContextFort

ContextFort provides visibility and controls for AI coding agents running on engineers' machines. When a developer runs a tool such as Cursor or Claude Code, the agent inherits that developer's system permissions, and ContextFort records what it does with them. It is aimed at security and engineering teams that want an audit trail of agent activity. The vendor contrasts it with traditional endpoint detection, which asks whether a process is malicious, whereas ContextFort asks what an agent accessed. It monitors from the operating system, independently of the agent, using eBPF on Linux, the Endpoint Security Framework on macOS, and ETW with a minifilter on Windows. Listed telemetry covers every file an agent reads or writes, including environment files and credentials, outbound network connections and the process that initiated them, and the full tree of subprocesses such as shell commands, package installs and builds. The vendor states the logs are kept independently so the agent cannot tamper with them. The site also lists sandboxing and platform areas alongside telemetry. Pricing is not published. Access is through a demo booking, and the site links to a GitHub repository. The vendor states that it is backed by Y Combinator.