LLM Clip
LLM Clip is built to capture browser evidence and hand it to Claude or Codex while keeping the trust model simple: clips stay local unless you explicitly send them.
When you trigger a clip, LLM Clip may capture a screenshot of the current tab or selected region, page metadata such as title and URL, selected text, a compact DOM summary, and runtime context such as console output, route changes, network request summaries, and a bounded debugger snapshot when Chrome allows it.
This information is captured to help you show a browser bug or UI issue clearly to an AI coding assistant.
Clip data is stored locally in your browser and extension storage. Image assets are stored locally. LLM Clip does not require an account for v1 and does not sync your clips to a remote service by default.
Data leaves your machine only when you take an explicit action, such as copying evidence to the clipboard, sending a packet through the local bridge to Claude or Codex, or using an optional future integration.
The localhost bridge is an optional companion that runs on your machine and can write deterministic bundles into your local workspace or resume a local Claude/Codex session.
The current v1 build keeps third-party integrations disabled. If future versions add optional integrations such as Slack, Jira, Discord, Linear, or Microsoft Teams, any data you choose to send through them will be governed by those services’ own terms and privacy practices.
LLM Clip does not sell your personal information. The product is designed around local capture and explicit handoff, not silent collection.