Live page evidence
The element, region, and visible state come from the interface you are actually reviewing, not a screenshot someone pasted into chat.
Agent Workflow · Browser Context
Give your coding agent the browser context it cannot infer.
Quey starts with the live interface, not a paraphrase of it. Select the UI, capture the change request, expose it to your local provider, and let the agent begin from a source-aware handoff instead of another round of file hunting.
Works with Codex, Claude Code, Cursor, ChatGPT, Windsurf, and other MCP-aware workflows.
What Changes
Agents stop guessing which part of the UI you meant.
Quey is not another chat wrapper. It is the missing browser-side evidence layer for UI edits that start from what someone actually selected on the page.
Source-aware handoff
Component clues, selector paths, and source hints reduce file hunting before the first line of code gets touched.
Reviewable workflow
A capture can be acknowledged, replied to, dismissed, or resolved so the request stays visible after the agent finishes.
From Page To Patch
A tighter loop from browser review to source edit.
The page tells you what changed. Quey tells the agent where that change came from, how it was requested, and how to report back when the work is done.
01
Select the UI
Pick the exact element or region that needs work in the browser.
02
Capture the request
Quey bundles the change request with layout, copy, and source hints.
03
Expose it to the agent
The local provider makes that capture available through HTTP and MCP inside the repo you want to edit.
04
Resolve with context
The agent ships the patch and writes back a summary instead of leaving the fix buried in chat history.
Why It Holds Up
A payload the repo can act on.
When an agent receives only prose, it has to reconstruct the UI problem from memory and probability. Quey shortens that path by attaching the request to the live page evidence that produced it.
That evidence is small enough to move quickly, structured enough to automate around, and explicit enough to survive a team workflow where designers, reviewers, and engineers are all touching the same request.
What a capture preserves
Enough context to make the first edit defensible, not lucky.
Target identity: selector path, element type, and visible copy
Source clues: file path hints, component name, and line hints when available
User intent: the request title, body, and change mode
Workflow state: pending, acknowledged, dismissed, or resolved
Surfaces
Use the layer you need, not a new stack.
Quey's agent story is split on purpose: local provider for project-aware edits, MCP for tool interoperability, and a schema contract when you need to wire the data into something custom.
CLI
Run the provider where the code lives.
Start the local provider in the repo you want to edit so browser captures land in the right project.
MCP
Use the same capture model across clients.
Codex, Claude Code, Cursor, ChatGPT, and other MCP-aware tools can all consume the same structured handoff.
Schema
Inspect the payload before you automate around it.
See the contract for what Quey exports when a user selects an element, region, or annotation.
The practical questions teams ask first.
Everything you need to know about Quey for agents
A capture includes the selected target, the user’s requested change, source-aware hints, and the visible browser context that led to the request. The goal is to start the edit from evidence instead of guesswork.
No. Quey is deliberately tool-neutral. The provider exposes the same capture loop through HTTP and MCP so you can plug it into whatever agent workflow your team already uses.
Yes. The extension captures context in the browser, and the local provider exposes that context inside the project you want the agent to edit.
Optionally. There is an auto-run path, but the safer default is a reviewable loop where the agent acknowledges the capture, makes the edit, and resolves it with a summary.
Start With The Page
Show the agent the interface before you ask for the fix.
Install the extension, run the provider inside the repo you want to edit, and let the request begin with selected UI instead of a blurry handoff in chat.
Better prompts matter. Grounded prompts matter more.
Quey captures live browser context and exposes it to coding agents through a practical local provider.