See every request your coding agent sends.

Dex is free for developers.

Works with Claude Code, Codex, GitHub Copilot, and more.

Dex, the robot mascot, waving
Dex, in dark mode, listing two conversations: a Codex question about Go versions, and a Claude Code changelog task with 7 blocked activity, above a timeline of model calls, tool calls, processes and requests

Every prompt, tool call, and request.

Including what it sends back to Anthropic, OpenAI, or GitHub.

A Claude Code conversation in Dex: a timeline of model calls, tool calls, processes, files and requests, token usage, and the start of its evidence list

Dex stops risky actions.

Your agent finds out why, and can ask you to approve.

Dex showing a blocked POST to api.github.com/user/repos from a trojaned gh command, judged likely harmful, with Archodex closing the connection before the request was sent

Your agent can't get around it.

Dex runs outside the agent and covers every process it starts.

Dex evidence for a Claude Code conversation, showing each Bash tool call and the gh processes it started, including /usr/local/bin/gh

Using Docker for sandboxing? Still covered.

Archodex automagically instruments Docker Desktop, auditing and protecting all AI assistant activity occurring inside your containers.

An exploded stack of layers. At the base, your laptop. Above it, Dex. Above that, Claude and ChatGPT running on the laptop, beside Docker Desktop. Above Docker Desktop, the Archodex Linux Agent, and on top, claude and codex running in containers. Lines run from Claude and ChatGPT down to Dex, from claude and codex down to the Archodex Linux Agent, and from the Linux Agent down to Dex.

Be first to try Dex.

For teams: AI Endpoint Protection