What is md-log?

md-log in one line

md-log is a review viewer and archive layer where an AI coding agent writes its work and analysis as Markdown (.md) reports, and a person comfortably reads, reviews, and approves those reports on web, mobile, or tablet. The agent saves reports by file path over MCP, reviewers annotate with a pen (S-Pen or Apple Pencil), and every save piles up as an immutable version.

The whole idea fits in one line: you review the REPORT, not the diff. md-log is not an automated code reviewer that judges your code for you — it is a review and record-keeping tool that supports human judgment.

The problem it solves — the vibe coding review bottleneck

In so-called vibe coding — where you steer an agent in natural language and it generates code in bulk — it is effectively impossible for a person to read every diff line by line. The volume is large, the context is scattered, and review is tied to the IDE at your desk.

As a result, reviews become perfunctory, and there's no lasting record of what the agent changed, why, and with what impact — so a teammate or a new agent later loses the context.

The solution — review the report, let a human decide

md-log moves the unit of review from the diff to the report. The agent leaves a Markdown report summarizing WHAT / WHY / IMPACT, and a person reads it and approves or rejects. The decision-maker is still the human (human-in-the-loop); md-log simply swaps the surface your eyes land on — from a pile of diffs to a readable document.

How it works, at a glance — 3 steps

1. The agent writes a report. MCP clients like Claude Code, Claude Desktop, OpenAI Codex, and Cursor connect by URL (https://mcp.md-log.com/mcp) — no install — and each time they finish a task they save a .md report by path. Folders are auto-created on save (like mkdir -p), and images upload too.
2. You review and annotate. Read the report on the web app, mobile, or tablet; on pen-capable tablets you hand-annotate with a stylus. See Review & annotation.
3. It accumulates as immutable versions. Every save becomes a new version, so your project's AI-collaboration history builds up — doubling as onboarding material for new teammates or new agents.

Three tracks — Individual, team collaboration, and On-Premise

md-log comes in three tracks.

	Individual	Team collaboration	On-Premise
Delivery	Operator-hosted SaaS, free	Operator-hosted SaaS, free	Self-hosted
People	1 person	Up to 5 members per team	No member limit
Getting started	Sign up at app.md-log.com → mint your MCP key in Settings → Tokens	Sign in, create a team, invite teammates by email	Contact sales to adopt
Where data lives	Operator's servers	Operator's servers	Behind your firewall, on your own infrastructure
Best for	Solo or small experiments, start instantly	A small team reviewing and annotating reports together	More members, or keeping reports carrying code context and decisions (IP) out of an outside cloud
Sign-up	No credit card, no sales inquiry	No credit card, no sales inquiry	Contact us

Team collaboration is a free feature inside the hosted SaaS, with up to 5 members per team. Share a document or a whole folder (its entire subtree) into a team, and members read the reports, editors and above save new versions, and everyone sees each other's pen annotations pinned to the version. For the details, see collaboration. When you need more members or want reports kept behind your own firewall, move up to On-Premise self-hosting.

Key concepts

• MCP auto-collection: Add one rule to your agent rules file (CLAUDE.md for Claude Code, AGENTS.md for Codex, project rules for Cursor) and the agent writes a report after each task — no need to say "save this" every time.
• Immutable versions: Every save, update, and append becomes a new version, and its commit_message shows up in the version history. Moving or renaming preserves history and annotations.
• Multi-device + stylus annotation: Review on web, iOS, and Android (phones and tablets); on pen-capable tablets, hand-annotate with an S-Pen or Apple Pencil. Annotations are pinned to a specific document version.
• Team collaboration (free): Create a team, invite teammates by email (up to 5 members per team), and share documents or folders to review and annotate together — shared reports show up in the web file browser and in a dedicated mobile tab. See collaboration.
• Project memory / onboarding: The reports that accumulate become your project's AI-collaboration history, helping new teammates and new agents get up to speed fast.

Who it's for

md-log is for developers, tech leads, EMs, and small teams running a plan → review → approve loop with an AI coding agent. It's especially useful for reviewers who want to vet an agent's work asynchronously — on a tablet, away from the desk.

Next steps

• Get started in five minutes → Quickstart
• Connect your agent to md-log → Connect MCP
• Learn about review and annotation → Review & annotation
• Review together as a team → Collaboration
• Try it now → app.md-log.com