The Vercel plugin on Claude Code wants to read your prompts
akshaychugh.xyz
April 9, 2026
6 min read
🔥🔥🔥🔥🔥
57/100
Summary
The Vercel plugin on Claude Code requests permission to collect anonymous usage data, including all prompt text, even in non-Vercel projects. The plugin assists with deployments, framework guidance, and skill injection.
Key Takeaways
- The Vercel plugin collects extensive telemetry data, including full bash command strings and prompt text, even for non-Vercel projects.
- The plugin injects behavioral instructions into Claude, allowing it to ask users for consent and execute commands based on their responses without clear visual indicators.
- Users are not adequately informed that their bash command collection is always on, and the consent framework is misleading, framing the choice as between sharing prompts or not.
- An opt-out option exists for telemetry, but it is not easily accessible, as it is documented only in the plugin's README within the cache directory.
Community Sentiment
Positives
- The Vercel team is actively improving the plugin and values user feedback, indicating a commitment to enhancing user experience and functionality.
- The plugin aims to support a wide range of projects, including greenfield applications, which could foster innovation in AI chat app development.
Concerns
- The default telemetry feature raises significant privacy concerns, as it collects full bash commands and sensitive information without user consent.
- Injecting skills into sessions unrelated to Vercel creates unnecessary costs and complexity for users working across diverse projects.
- There is a breach of trust due to the plugin's behavior, which violates established policies and could lead to security vulnerabilities.