Click on UI elements, create threads, attach screenshots and audio. While you keep reviewing the interface, your AI agent works in parallel — picking up threads, commenting, and resolving issues without interrupting your flow.
A continuous loop between humans and AI, centered on your actual interface.
Toggle the SpecLens switch in any app with the script installed. Every element with a data-spec-id becomes clickable. You can also use Ctrl+Shift+S.
Click the element you want to discuss. SpecLens identifies it by its spec ID and highlights it with the amber outline.
Write your annotation. Describe the issue, request a change, or ask a question. Each thread is tied to that specific UI element.
Attach a screenshot directly from the overlay. Record audio that gets auto-transcribed by AI. Your agent sees images, transcriptions, and text — full context.
While you keep reviewing the interface, your AI agent picks up pending threads. It comments before and after each change, keeping the conversation alive. You don't need to stop your flow.
The AI comments its resolution on the thread and marks it as resolved. You see the green checkmark. The conversation stays as a record of what was discussed and done.
SpecLens has two sides: one for your app, one for your AI.
Add the overlay to any web application with a single script tag.
Give your AI agent a prompt and API access. It handles the rest.
Two steps to close the loop between your team and your AI.
Add this script tag to any page. Set your app ID, screen name, and token.
Then add data-spec-id="element.name" to elements you want to annotate.
Add this to your AI agent's system prompt or Claude preferences. It tells the agent how to consume and resolve threads.
Thread-centric API. Auth via X-Spec-Token header.