Skip to main content
Glama

@kernline/mcp

Local MCP server for the Kernline browser extension. It lets Cursor, Claude Desktop and other MCP clients read inspection output (CSS, Tailwind, design tokens, Copy for AI bundles) from tabs where the Kernline panel is open.

Public source for this package is the read-only mirror devson-software/kernline-mcp. Kernline copies this folder there on push. Do not edit the mirror.

Everything stays on your computer. The server listens on 127.0.0.1 only, accepts connections from the Kernline extension only, and proves itself with a pairing token before any data is shared. Kernline never receives page content on its servers. Your AI app may send what it reads to its own model provider.

Setup

  1. In Kernline, open Settings, go to MCP server, tick Enable MCP server, then click Copy token.

  2. Add the server to your MCP client with that token.

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "kernline": {
      "command": "npx",
      "args": ["-y", "@kernline/mcp"],
      "env": { "KERNLINE_BRIDGE_TOKEN": "paste-your-token" }
    }
  }
}

Claude Desktop (claude_desktop_config.json)

Use the same kernline block under mcpServers.

  1. Open the Kernline panel on a tab, then ask your AI app to use the Kernline tools.

Related MCP server: MultiZen

Options

  • KERNLINE_BRIDGE_TOKEN (required) is the pairing token from Kernline settings.

  • KERNLINE_BRIDGE_PORT (optional, default 47321) must match the port in Kernline settings.

Requires Node.js 18 or newer. Pro tools (Copy for AI, page tokens) need a Kernline Pro licence in the extension.

License

MIT, made by Devson Software in South Africa. The Kernline browser extension itself is not open source.

Available Tools

10 tools
kernline_assetA
Read-only

Fetch one asset by id from kernline_assets (inline SVG or text, size-capped)

Inputs: id (required), tabId.

kernline_assetsA
Read-only

List page assets discovered by Kernline (images, SVG, fonts, favicons)

Inputs: tabId.

kernline_component_captureB
Read-only

Pro: component tree capture as html, react, or vue with Tailwind classes

Inputs: tabId, format.

kernline_copy_for_aiA
Read-only

Pro: copy-for-AI bundle (HTML, computed CSS, tokens, sizes, states) for the pinned element

Inputs: tabId.

kernline_element_imageA
Read-only

PNG screenshot crop of the pinned element or selector (requires Kernline open on tab)

Inputs: tabId, selector.

kernline_inspectA
Read-only

Computed style snapshot, cleaned CSS, and Tailwind classes for the pinned element or a CSS selector

Inputs: tabId, selector.

kernline_list_tabsA
Read-only

List browser tabs where Kernline is open and connected to the local bridge

Inputs: no named parameters.

kernline_page_summaryC
Read-only

Viewport, frames, and blocked stylesheet origins for an inspected tab

Inputs: tabId.

kernline_page_tokensC
Read-only

Pro: scan page design tokens (css variables, tailwind v3/v4, json, or AI context)

Inputs: tabId, format.

kernline_pinB

Scroll a selector into view and pin it in Kernline so the user sees what was selected

Inputs: selector (required), tabId.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.5/5.0

    Scored across 10 tools

    Disambiguation4/5

    Most tools have clearly distinct purposes: tab listing, page summary, asset listing/fetching, screenshot capture, and pinning are well separated. The main ambiguity is among inspect, copy_for_ai, component_capture, and element_image, which all operate on a pinned element or selector, though their output formats are described differently.

    Naming Consistency4/5

    All tools use a consistent kernline_ prefix and snake_case naming, making the set easy to scan. However, the naming mixes verbs and nouns (e.g., kernline_inspect, kernline_pin vs. kernline_page_summary, kernline_assets) rather than following a strict verb_noun pattern.

    Tool Count5/5

    Ten tools is well-scoped for a browser/design inspection bridge, falling comfortably within the typical 3-15 range. Each tool appears to cover a distinct inspection or capture capability without excessive surface area.

    Completeness4/5

    The surface covers the core lifecycle well: listing tabs, inspecting pages/elements, capturing styles/components/tokens, listing and fetching assets, screenshotting, and pinning elements. Minor gaps exist, such as no explicit unpin/clear-selection operation and limited asset fetching for binary files, but these are likely workable.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      A
      quality
      B
      maintenance
      Lets an AI share the same live browser tab as a person, receiving the user's Alt+click target and enabling the AI to inspect, click, fill, type, navigate, or otherwise operate on that real page. It runs locally and pairs with the browser extension to bridge the user's page to the MCP client.
      13
      -
    • F
      license
      Not graded
      quality
      A
      maintenance
      Exposes a localhost MCP endpoint that lets AI agents drive isolated browser profiles — navigating, clicking, typing, extracting, screenshotting, evaluating JS, and managing tabs and cookies — so real authenticated workflows can be completed from MCP clients like Cursor or Claude Desktop. Cookies, login state, fingerprints, and proxies persist per profile across launches and AI sessions.
      -
    • F
      license
      Not graded
      quality
      A
      maintenance
      Connects web-based AI chat clients such as Claude, ChatGPT, and Grok to a local machine through a single OAuth-gated MCP endpoint so they can browse, read, write, and edit files, run whitelisted shell commands, search content, use Git, query SQLite, drive browsers and DevTools, and manage background tasks, ports, and clipboard. It also exposes localhost and LAN services via a protected HTTP client, letting remote or mobile clients debug and operate local projects without granting unrestricted shell access.
      -
    • A
      license
      Not graded
      quality
      A
      maintenance
      A local MCP server that lets AI coding tools drive the user's actual, already-logged-in Chrome browser — snapshotting pages, clicking, typing, navigating, reading content, taking screenshots, and inspecting cookies, network traffic, and console output. Browser automation happens through a Chrome extension using Native Messaging, so no remote-debugging port is exposed and all communication stays on the local machine.
      1 npm
      ISC