Integrations
The AI tools Evidren works with, and what it can do with each.
AI coding agents
With Claude Code, Evidren checks shell commands, file writes and edits in the folders you connect, before they run. File reads, web fetches and other tools are not checked before they run.
| Question | Claude Code | GitHub Copilot CLI | OpenAI Codex CLI |
|---|---|---|---|
| See what the agent is about to do, and write it down | Yes | Yes | Not verified |
| Stop it before it runs | Yes | Yes | Not verified |
| Hand the decision to you | Yes | Yes | Not verified |
| Replace a credential on the way through | Yes | No | Not verified |
| Run against the real tool | 24 October 2026 | 24 October 2026 | 14 September 2026 |
OpenAI Codex CLI: OpenAI’s coding agent. Evidren has mapped it but cannot cover it yet.
AI chats in Chrome and Edge
The browser extension checks messages, pasted text and files you send to AI chats, before they leave the browser. When it finds personal data or an access key, it asks first and can mask it.
What Evidren does not see, and does not stop
An open Evidren window does not mean your computer is protected. Evidren covers the tools it is connected to, in the folders you connected, and nothing else. If an action never reaches Evidren it leaves no line, and a missing line cannot be spotted afterwards.
- Blocking is best effort.
- The rules match the shape of a command, so a determined adversary can get past them. Measured against an adversarial set of 81 shell commands: 57 handled, 24 not.
- It can be switched off.
- One line in an agent's own settings turns every hook off. Evidren detects that and reports it. It cannot prevent it, and it cannot promise to detect every way of disconnecting it.
- The record proves what is in it.
- The chain shows that the lines present were not altered. It cannot show you a line that was never written.