Snapstack

@bgaze

5 days ago

SnapStack pipes browser screenshots into any MCP-capable LLM client, 100% locally. Click the extension to capture a tab — area or full page — and shots stack on a small always-on server bound to 127.0.0.1. Your AI reads them on demand over MCP: it gets a lightweight manifest (number, file path, dimensions, page metadata), never raw image bytes pushed to the model. No account, no telemetry — nothing ever leaves your machine.

Overview Tools Comments

Overview

SnapStack

Please see for more details : https://github.com/bgaze/snapstack-extension

Capture any browser tab in one click and stack it locally — your AI reads the screenshots on demand.

SnapStack lets you hand browser screenshots to your AI assistant without copy-pasting images. Click the extension, your shot is stacked locally; ask your LLM to "look at my screenshots" and it picks them up over MCP. It works with any MCP-capable client (Claude Code and others).

SnapStack is 100% local. Captures go only to a small server on your own machine (127.0.0.1). Nothing is ever uploaded, no account, no telemetry. See PRIVACY.md.

SnapStack has two halves: this extension and a small always-on local server that holds the stack and serves it to your LLM. Both install in a couple of minutes.

Try in Playground

Server Config

{
  "mcpServers": {
    "snapstack": {
      "type": "http",
      "url": "http://127.0.0.1:4123/mcp"
    }
  }
}

Build with ShipAny.