Security

Last updated: May 25, 2026

Security is a core design goal of OpenConduit, not an afterthought. This page describes how we protect your data and how to report a vulnerability.

Found a vulnerability?

Use the form at the bottom of this page or email [email protected] directly. Do not open a public GitHub issue for security vulnerabilities. We will respond within 72 hours and credit you in the release notes once a fix is shipped.

1. API keys and credentials

API keys are stored in your operating system's native app-data directory (electron-store), which is accessible only to your user account. They are never logged, never included in crash reports, and never transmitted to any OpenConduit server. Keys only leave your device as Authorization headers sent directly to the provider endpoint you have configured.

2. No OpenConduit middleman

Chat messages travel directly from your device to the AI provider API. OpenConduit does not operate a proxy, logging service, or message store between you and the provider. The only exception is Live Collaboration (a Labs feature), where messages are relayed through a Cloudflare Durable Object to other session participants — and only for the duration of the active session.

3. Process isolation

OpenConduit runs in Electron with contextIsolation: true. The renderer process has no direct access to Node.js APIs or the file system. All privileged operations (provider calls, file access, settings reads/writes) go through a narrow IPC channel exposed via a contextBridge preload script.

4. MCP tool approval

When an AI model requests a tool call via the Model Context Protocol, OpenConduit shows an approval prompt before executing it. You can require explicit approval for every tool call in Settings → Require Tool Approval. No tool runs silently without your knowledge.

5. Automatic updates

The auto-updater fetches release metadata from updates.openconduit.ai over HTTPS. On macOS and Windows, update packages are code-signed. You can disable automatic update checks in Settings.

6. Open source

The full application source code is published at github.com/OpenConduit/Client under AGPL-3.0. Anyone can audit, fork, and verify the build. We encourage security researchers to review the code and report any issues through the responsible disclosure channel above.

7. Responsible disclosure policy

  • Use the form below or email [email protected] with a description, reproduction steps, and impact assessment.
  • We will acknowledge receipt within 72 hours.
  • We will work with you on a fix and target a patch release within 14 days for critical issues.
  • We will publicly credit you in the release notes unless you prefer to remain anonymous.
  • We ask that you do not publicly disclose the issue until a fix is released.

8. Scope

The following are in scope for responsible disclosure:

  • OpenConduit desktop application (macOS, Windows, Linux)
  • openconduit.ai website and Cloudflare Worker infrastructure
  • Live Collaboration relay (Cloudflare Durable Object)

Out of scope: vulnerabilities in third-party AI provider APIs, Electron itself, or your operating system.

Report a vulnerability

Use this form to submit a security report privately. Do not include exploit code or live credentials.