Markloop is a web app for collecting human feedback on HTML documents created by AI coding agents, then packaging that feedback into structured context the agent can consume over MCP or as plain files.
What is Markloop?
Markloop is a feedback tool built specifically for agent-made HTML artifacts — specs, reports, proposals, RFCs, and other documents. You upload or push an HTML file (via MCP or the browser), share a rendered version with reviewers, collect pinned comments and answers, and export the feedback as an agent-ready package. No changes run on Markloop’s servers — your agent applies edits locally and publishes the next version. The product is made by a small team (founder-led) and offers a 14-day free trial with no credit card required.
Key Features
- Anchored comments — Reviewers pin a comment to a specific element and exact text quote; comments stay attached to the version they were made on.
- Agent-ready feedback packages — Every open comment exports with its CSS selector, quote, reviewer intent, surrounding context, and version number — over MCP (native for Claude Code and Codex) or as a plain markdown file (
comments.md).
- MCP integration — Native MCP tools for Claude Code and Codex: your agent can push a new version (
markloop_put_version) and pull open feedback without leaving the terminal.
- Versions with addressed tracking — Each upload becomes a version in a chain; resolved comments are marked ✓, open questions carry into the next round.
- Unlimited reviewers on every plan — Invite clients, teammates, or stakeholders as Viewers; they see the rendered document (not the source HTML) and can comment but not edit.
- Read-only file sharing — Share a single document via private or public link with optional expiry and version-access controls; no account needed for readers.
- Questions & answers — Embed yes/no or open-ended questions directly in the document; answers come back to your agent with the rest of the feedback.
Who is it for?
- Product & founders — Share a PRD or spec drafted by an agent, collect stakeholder sign-off, and let the agent fold comments into the next version.
- Engineering teams — Review RFCs, architecture docs, and migration plans with line-level comments; track what’s addressed across versions.
- Consultants & agencies — Send polished HTML reports and proposals to clients, collect feedback in context (no Slack threads or screenshots), and export the review for an agent to apply.
- Anyone with an AI coding agent — If your agent writes HTML artifacts (Claude Code, Codex, or any other tool), you can loop human feedback back into the agent’s workflow.
What can you do with Markloop?
- Get sign-off on agent-drafted specs — Upload a PRD, invite product stakeholders as Viewers, collect comments and answers, then pull the structured feedback into your agent via MCP — no copy-pasting.
- Review technical design docs in context — Share an RFC or architecture doc, pin comments to exact code blocks or diagrams, and track what’s been addressed as you iterate through versions.
- Loop client feedback on polished deliverables — Send a read-only link to a client report (e.g., SEO audit), let them comment and answer embedded questions, then have your agent pull the changes and publish v2.
How does Markloop work?
- Add — Upload an HTML file via the browser or push it from your agent over MCP (
markloop_put_version).
- Share — Invite Viewers to a project or share a read-only file link with expiry and version visibility controls.
- Collect — Reviewers pin comments and answer embedded questions; you see all feedback anchored to the document.
- Pull — Your agent pulls open comments over MCP as structured context (selector, quote, intent, version).
- Apply — Your agent applies changes locally, publishes a new version, and resolved comments are marked ✓. The loop repeats.
