Skip to content
MCP server

MarkLayer

Let AI coding agents work through visual annotations on a live webpage

Runs locallyStatus and history
Pricing
Open source
Platforms
CLI, Web, Chrome extension
Runs
Locally
Version
0.6.1
Listing updated
3 days ago

About MarkLayer

MarkLayer is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.

Reviewers mark up a page in MarkLayer and an agent sees those comments as a work queue. It can acknowledge a comment, make the fix, resolve it with a summary, dismiss it with a reason, or reply with a question, and every status change shows to the people in the room as it happens. The flow also runs the other way: point an agent at a page and it can read headings, text, links and controls, then leave its own comments or suggest exact copy edits as diffs.

Beyond single comments, tools let the agent open a new review room on a preview URL it built, publish an ordered checklist of next steps tied to annotation ids, and put a 2 to 6 option poll to the room. A watch call blocks until new annotations arrive, so an agent can loop on incoming feedback.

The npm package is the stdio build. The README adds that every MarkLayer share link is already a remote MCP server at the same address plus /mcp, bound to its room. MarkLayer itself is free, anonymous, Apache-2.0 and can be self-hosted; shared annotations are deleted 90 days after last opened.

Install

Run in a terminal

claude mcp add marklayer -- npx -y marklayer-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "marklayer": {
      "command": "npx",
      "args": [
        "-y",
        "marklayer-mcp"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "marklayer": {
      "command": "npx",
      "args": [
        "-y",
        "marklayer-mcp"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "marklayer": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "marklayer-mcp"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "marklayer": {
      "command": "npx",
      "args": [
        "-y",
        "marklayer-mcp"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "marklayer": {
      "command": "npx",
      "args": [
        "-y",
        "marklayer-mcp"
      ]
    }
  }
}

Configuration

  • MARKLAYER_ROOMSame as --room: the share URL or room id to join at startup.
  • MARKLAYER_API_BASEOverride the MarkLayer worker base URL. Defaults to https://marklayer.app; useful against a local dev worker.
  • MARKLAYER_AGENTName shown to the humans in the room for this agent. Defaults to claude-code.

Tools this server provides (17)

Taken from the project documentation; the running server may differ.

marklayer_connect_room
Connects to a room by share URL or bare id.
marklayer_room_info
Page URL, viewport width and timestamps.
marklayer_read_page
Page headings, text, links and controls with selectors.
marklayer_list_annotations
Lists annotations, optionally by status.
marklayer_get_annotation
Full detail and reply thread for one annotation.
marklayer_watch_annotations
Blocks until new annotations arrive and returns a batch.
marklayer_acknowledge
Marks an annotation in progress and tags this agent.
marklayer_resolve
Marks resolved, optionally with a summary reply.
marklayer_dismiss
Marks dismissed with a reason.
marklayer_reply
Posts a reply without changing status.
marklayer_create_annotation
Leaves a new comment.
marklayer_suggest_edit
Proposes an exact text fix as a diff.
marklayer_share_preview
Opens a new review room on a URL.
marklayer_set_next_steps
Publishes an ordered checklist linked to annotation ids.
marklayer_complete_next_step
Ticks a step off.
marklayer_ask
Puts a 2 to 6 option poll to the room.
marklayer_get_feedback
Reads feedback in a single call.

What it does

Annotation work queue
Agents list, read and watch annotations and their reply threads.
Visible status changes
Acknowledge, resolve and dismiss updates appear live to reviewers.
Page review
Reads page text and controls so an agent can review unannotated pages.
Suggested edits
Proposes exact copy fixes as diffs for the human to accept.
Review rooms
Opens new rooms on preview URLs and versions them by repo and branch.
Steps and polls
Publishes a next-step checklist and asks the room multiple-choice questions.

Good for

  • Fix every client comment on a staging page, then resolve each with a note
  • Have an agent copy-edit a live page and leave suggestions for approval
  • Track follow-up work as a checklist linked to feedback
  • Ask reviewers to choose between design options

Pricing

Open source

Free, no account; Apache-2.0 and self-hostable; optional one-off tip

Prices change. Check marklayer.app before you buy.

MarkLayer FAQ

Do I have to install anything?

Not necessarily. Every share link works as a remote MCP server at its address plus /mcp.

How do I identify which agent is acting?

Set MARKLAYER_AGENT or the agent query parameter; humans see it as the comment author.

What does MarkLayer cost?

Nothing. The pricing page says there is no paid plan, trial or usage cap.

How long are annotations kept?

Shared annotations are deleted 90 days after they were last opened.

Is MarkLayer free?

Yes. MarkLayer is open source, so you can use and self-host it for free.

How do I add MarkLayer to Claude Code?

Run this command in your terminal: claude mcp add marklayer -- npx -y marklayer-mcp

What are the best alternatives to MarkLayer?

The closest alternatives to MarkLayer are REA, HyperFrames by HeyGen, Codebase Memory, GitHub and Serena MCP: the IDE for your agent.

Alternatives to MarkLayer

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Hosted MCP connector for creating and rendering videos from HTML through AI chat

    • Remote
    • Freemium
    • 85.3K GitHub stars
    • Design and media
  3. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  4. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  5. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  6. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory