Visual feedback for AI coding agents

Your AI can't see
the screen.

Screenshots vanish in the terminal. BlindSpot bridges the gap — your AI uploads, you annotate on your phone, and the AI fixes the code.

1

AI captures

Your agent takes a screenshot and uploads it through BlindSpot's MCP server.

2

You annotate

Open the URL on your phone. Draw, circle, arrow, and comment with touch.

3

AI fixes

The agent reads your annotations — shapes, coordinates, comments — and updates the code.

The Problem

You're on your phone, remote-controlling Claude Code. You tell it to fix a layout bug. It takes a screenshot… but you can't see it. The terminal doesn't render images. You're debugging blind.

The Solution

BlindSpot gives your screenshots a shareable URL. Open it on your phone, draw circles and arrows, add comments like "move this left". Your AI reads the structured feedback and fixes the code.

Workflow

Three steps, zero context switching

1

AI uploads a screenshot

Your AI agent captures the screen via Playwright or any screenshot tool, then uploads through BlindSpot's MCP server. A secure, temporary URL is returned.

"Take a screenshot and upload it for me to review"
2

You annotate on your phone

Open the URL on your phone browser. Draw arrows, circles, rectangles with touch. Add comments to each shape — "change this color", "remove this button". Touch-optimized editor, zero learning curve.

3

AI reads and fixes

Your agent calls get_annotation to read structured data — shape types, coordinates, and your comments. It knows exactly what to change.

Capabilities

Everything your AI needs to see

Touch Annotation

Arrows, circles, rectangles, freehand. Add text comments to any shape.

MCP Integration

17 tools out of the box. Claude Code, Claude Desktop, any MCP agent.

Visual Diff

Compare before/after screenshots. Changed pixels highlighted automatically.

Programmatic

AI draws boxes, arrows, numbered steps directly on images.

Auto-Expiry

Screenshots auto-delete. Private bucket with signed URLs only.

Zero Setup

Add MCP config, get API key, done. No extensions needed.

MCP Server

17 tools, one install

Your AI gets all of these the moment you add the MCP config.

Upload & Annotate
upload_screenshotget_annotationlist_screenshots
Draw on Images
annotatebatch_annotateannotate_stepreset_callout_counter
Transform
blur_regioncrop_imageresize_imagevisual_diff
Manage
load_imagesave_imageget_imagelist_loaded_imagesdelete_loaded_imageundo

Get Started

Two minutes to set up

claude-code-config.json
{
  "mcpServers": {
    "blindspot": {
      "command": "node",
      "args": ["path/to/blindspot/mcp-server/dist/index.js"],
      "env": {
        "VISUAL_FEEDBACK_URL": "https://kortress.com",
        "VISUAL_FEEDBACK_API_KEY": "your_api_key"
      }
    }
  }
}

Then just tell your AI: "Upload a screenshot for me to annotate"

Pricing

Start free, scale when ready

No credit card required. Upgrade anytime.

Free

$0 /month

  • 100 screenshots / month
  • 1 hour retention
  • 1 API key
  • All 17 MCP tools
  • Touch annotation editor
Popular

Plus

$3.99 /month

  • 1,000 screenshots / month
  • 24 hour retention
  • 3 API keys
  • All 17 MCP tools
  • Touch annotation editor
Upgrade to Plus

Stop coding blind.

Set up in 2 minutes. Free forever for personal use.