MarkLayer — Claude Code MCP server: features and setup
MarkLayer brings webpage feedback into an agent workflow, where the agent can reply, update statuses, and suggest edits. Use it to review a page with other people.
claude mcp add --env MARKLAYER_ROOM=<MARKLAYER_ROOM> --env MARKLAYER_API_BASE=<MARKLAYER_API_BASE> --env MARKLAYER_AGENT=<MARKLAYER_AGENT> --transport stdio mcp -- npx -y marklayer-mcpWhat MarkLayer does
The server connects MarkLayer page annotations to an agent. It can mark a comment as in progress, reply, close or dismiss it. It can also inspect a page before anyone has commented, add its own feedback, and propose a precise text change for a person to review.
Tools cover rooms, annotations, replies, polls, and ordered next steps. The agent can wait for incoming notes and update their status for everyone in the room to see. The stdio setup runs through npx; provide a room at startup or connect to one later. A first-time npx -y download can outlast the MCP client's startup timeout, so the initial connection may fail.
What it helps with
- process new feedback on a webpage
- mark comments as in progress or resolved
- suggest a focused copy edit for page text
- publish a checklist tied to annotations
- collect reviewer votes in a room
Example prompts for the agent
- Watch the MarkLayer room, fix each new comment, then close it with a short summary.
- Read the page in this room and leave feedback on its interface.
- Suggest a grammar correction for the page and ask the reviewers to choose an option.
How to install MarkLayer
Platform CLI
claude mcp add --env MARKLAYER_ROOM=<MARKLAYER_ROOM> --env MARKLAYER_API_BASE=<MARKLAYER_API_BASE> --env MARKLAYER_AGENT=<MARKLAYER_AGENT> --transport stdio mcp -- npx -y marklayer-mcpReplace <…> with your own values
Platform CLI
codex mcp add mcp --env MARKLAYER_ROOM=<MARKLAYER_ROOM> --env MARKLAYER_API_BASE=<MARKLAYER_API_BASE> --env MARKLAYER_AGENT=<MARKLAYER_AGENT> -- npx -y marklayer-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"mcp": {
"command": "npx",
"args": [
"-y",
"marklayer-mcp"
],
"env": {
"MARKLAYER_ROOM": "<MARKLAYER_ROOM>",
"MARKLAYER_API_BASE": "<MARKLAYER_API_BASE>",
"MARKLAYER_AGENT": "<MARKLAYER_AGENT>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"marklayer-mcp"
],
"env": {
"MARKLAYER_ROOM": "<MARKLAYER_ROOM>",
"MARKLAYER_API_BASE": "<MARKLAYER_API_BASE>",
"MARKLAYER_AGENT": "<MARKLAYER_AGENT>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e MARKLAYER_ROOM=<MARKLAYER_ROOM> -e MARKLAYER_API_BASE=<MARKLAYER_API_BASE> -e MARKLAYER_AGENT=<MARKLAYER_AGENT> mcp npx -y marklayer-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about MarkLayer
How do I install MarkLayer in Claude Code?
Run claude mcp add marklayer -- npx -y marklayer-mcp. To connect a room at startup, use the separate setup command with the room and agent parameters.
What does MarkLayer do?
It brings webpage comments to an agent and provides tools to reply, change their status, or leave new notes. The agent can also suggest a specific text edit for human review.
Do I need to connect to a room separately?
You can supply a room at startup with --room, or connect later using marklayer_connect_room. A remote MCP endpoint from a room link is already tied to that room and cannot connect to another one.
