md-log-introduction-and-user-guide

Colin · 7/17/2026

md-log — Read and question AI-written code from anywhere, before you approve it

Review starts with the report, not the diff. md-log is a review viewer and archive layer where the work and analysis reports left behind by AI coding agents are comfortable to read, review, and approve on web, mobile, and tablet — and where every record accumulates as an immutable version.

What is md-log?

Vibe coding moves fast. The agent writes the code, and the code lands — but the reasoning behind it evaporates. Scrolling a thousand-line diff on a laptop isn't review, and it's definitely not something you can do on a phone.

When you build with an AI agent, the agent produces far more than code: analyses, decisions, trade-offs, and things it tried and discarded. Almost all of it disappears the moment the session ends. md-log gives that work a place to live.

md-log moves the unit of review from the diff to the report. The agent leaves a markdown report summarizing what it changed, why, and what it affects; you read that report and approve it or send it back. The judgment stays with the human (human-in-the-loop), and md-log changes the surface your eyes land on from a pile of diffs into a document that's actually readable.

md-log is not an automated code reviewer. It doesn't grade your agent's output. It makes that output reviewable by a person.

How it works, in three steps

Key features

• 🧩 Connect over MCP, nothing to install — point your client at the remote URL (https://mcp.md-log.com/mcp) and attach your personal key (mdlog_pat_...). That's it. Your agent saves, updates, and appends to .md files by path, with version history, full-text search, and image attachments.
• ✍️ Review anywhere, annotate with a pen — reports are meant to be read on the couch, not squinted at in a terminal. Read on web, mobile, or tablet while you're on the move, and mark up the report the way you'd mark up paper: S-Pen (Android), Apple Pencil (iOS), or finger-draw where there's no stylus. Annotations are pinned to a specific version, and you can see each teammate's handwriting together.
• 🗂️ Immutable version archive — every save adds a new version that can't be rewritten, so "what changed last week, and why" stays exactly as it was recorded.
• 👥 Team collaboration — share a single document or an entire folder with your team, then read and annotate together on the same version. Permissions split into Viewer (read-only) and Editor (save new versions, annotate, create documents).
• 🔒 Control of your data — reports, images, and annotations don't leak outward; they stay inside your organization's storage. And if you want them inside your own company infrastructure, you can run md-log self-hosted (On-Premise).

Document visibility — private by default, public when you want it

Every document starts private. You decide, per document, how far it travels.

Publish a document publicly to the web and anyone can read it at a public URL (md-log.com/p/…) without logging in. The version is pinned at the moment you publish, so keep editing privately afterward and the public copy stays exactly where you left it — you decide when readers see the next cut. The owner can unpublish at any time, and can choose whether search engines index the page. In other words: keep your team's internal information safely private, and open only the documents you want to share to the outside world.

Getting started — about 5 minutes

1. Sign up — create an account at app.md-log.com.

2. Issue an MCP key — mint a personal key yourself under Settings → Tokens. No request form, no waiting; copy it and use it right away.

3. Connect your agent — register the remote URL and your key in your MCP client, like this:

   {
     "mcpServers": {
       "md-log": {
         "type": "http",
         "url": "https://mcp.md-log.com/mcp",
         "headers": { "Authorization": "Bearer mdlog_pat_your-token-here" }
       }
     }
   }
   

4. Review — ask your agent to file a report (something like "Summarize what you just changed and save it to MyProject/2026-07-17-auth-refactor.md in md-log, with a short commit message" — the folder is created for you), then read it on web, mobile, or tablet and approve it, send it back, or annotate it with a pen.

Three tracks — Individual, Team, On-Premise

Learn more — the md-log user guide

From the overview through the 5-minute quickstart, connecting an AI agent (MCP), reviewing and annotating reports, team collaboration and On-Premise, and the FAQ — the user guide gathers everything you need in one place, whether you're just starting out or rolling md-log out to a team. Reach for it whenever you want step-by-step instructions.

👉 User guide: md-log.com/en/docs 👉 Connecting an AI agent, in detail: md-log.com/en/docs/connect-mcp

About this document

This document is itself a report: written and saved into md-log over MCP, filed by path, and versioned there like any other report in the archive. Publishing is the feature that turns a report like this one into a public page anyone can open at md-log.com/p/… — no login, pinned to the version that was published, unpublishable at any time. Write it once for your team; publish it when it's worth showing. Same document, same history. You just chose who gets to see it.