Markloop is a review platform for agent-made HTML documents, enabling human feedback and agent-ready packages.
Submit your website to get discovered by thousands of potential customers and boost your SEO.
Get ListedMarkloop is a web application designed to close the feedback loop between AI coding agents and human reviewers. When agents produce HTML documents such as specs, reports, proposals, or technical docs, Markloop provides a structured environment where people can comment, answer embedded questions, and track revisions. The platform then packages all feedback into a machine-readable format that agents can consume directly via MCP or as plain files. Built for teams using Claude Code, Codex, or any other agent that outputs HTML, Markloop solves the problem of scattered feedback across Slack, email, and Google Docs.
Anchored Comments: Reviewers can pin comments to specific sections or sentences within the document. Each comment stays attached to the exact version it was made on, preserving context across revisions. Comments include a CSS selector, quoted text, and the reviewer's note.
Agent-Ready Feedback Package: Every comment is exported with its target selector, quote, surrounding context, and version number. The package can be pulled over MCP by Claude Code or Codex, or downloaded as a plain comments.md file for any other agent.
Version Tracking with Resolution Status: Each uploaded HTML file becomes a version in a chain. Comments are marked as addressed, open, or resolved as new versions are published. The system tracks which comments have been handled and which remain open.
Unlimited Free Reviewers: Project Viewers can be invited at no cost. They see the rendered document in the browser, cannot download the source HTML, and can only comment and answer questions. There is no per-reviewer charge.
MCP Integration for Claude Code and Codex: Agents can push new versions and pull feedback directly from the terminal using the Model Context Protocol. Commands like markloop_put_version and markloop_pull_feedback eliminate copy-paste.
Read-Only Sharing with Access Controls: Individual files can be shared via private or public links with optional expiry dates and version visibility restrictions. Reviewers see only what is shared.
Users upload an HTML document produced by an AI agent, either manually or via MCP. They then invite reviewers to a project or share a read-only link. Reviewers pin comments and answer embedded questions directly on the document. The agent pulls the structured feedback package, applies changes locally, and publishes a new version. Markloop tracks which comments are addressed and which remain open for the next loop.
Markloop is built for developers, product managers, consultants, and founders who use AI coding agents to generate HTML documents such as specs, RFCs, reports, proposals, and technical documentation. It suits solo builders and small teams working with Claude Code or Codex, as well as agencies that need to collect client feedback on agent-produced deliverables. The platform is designed for self-contained HTML artifacts, not live websites or multi-page applications.
