A tool for developers running a plan → review → approve loop with their AI coding agent — Claude Code, Claude Desktop, Codex, Cursor, and more. Review the work and analysis reports your agents leave over MCP on web, mobile, and tablet (with a pen), and keep every record as an immutable version.
Mint your MCP key yourself in Settings after signing up — no sales call needed.
See the real flow — reading the reports your agents leave, marking them up with a pen, and archiving each as an immutable version.
Read and review the work and analysis reports your AI agents leave — from anywhere — and pile every record up as immutable versions and history.
One rule line in your agent's config and Claude Code · Codex · Cursor leave their work and analysis as .md reports with no extra prompting. Folders organize by path, so nothing gets lost or mixed up.
Instead of drowning in a hundreds-of-lines diff, read the report the agent wrote and approve or reject. Make the report the entry point for human-in-the-loop review.
Review on web, mobile, and tablet while you're on the move. On a tablet, mark it up in your own handwriting with a stylus.
Every save is an immutable version. The .md files that used to scatter now accumulate instead of disappearing, so you can retrace past work, reviews, and approvals at a glance.
The records you pile up become your 'history of collaborating with AI'. New teammates and new agents catch up on project context fast.
Documents are private by default, but you can pick a specific one and publish it to a public URL that anyone reads without signing in — a channel for team know-how or external info. The published version is pinned, and you can unpublish anytime.
Work reports carry your company's code context and decisions. The MCP key (PAT) is document-only, so a key alone can't manage the account — and teams can keep everything behind the firewall with On-Premise.
A three-step loop: the AI writes the work report, a human reviews it comfortably, and that record naturally becomes project memory.
Your AI coding agent leaves its work and analysis as .md reports in md-log over MCP. Folders are created automatically by path.
Read the report on web, mobile, or tablet, approve or reject it, and leave pen annotations when you need to.
Every review and task stays as an immutable version and history — becoming your project memory as-is.
https://mcp.md-log.com/mcp — point Claude Code · Claude Desktop · Codex · Cursor at that URL with your key as a Bearer header, and their work reports show up where you can review them. No install, no Node.js. Mint the key in Settings after signing up — no build, no sales call. (Prefer a local process? `npx -y md-log-mcp` still works.)
Share a single report or a whole folder with your team, then read the same version together and mark it up by hand — on web, phone, or tablet.
Share one document or an entire folder (its whole subtree) with your team in one action — and make it private again anytime.
Invite a teammate who already has an md-log account by email, then split access: Viewer (read-only) vs Editor (save versions, annotate, create docs).
Shared items collect in a ‘Shared with me’ group in your file browser — on both web and mobile — each marked with a badge.
Everyone sees per-author pen annotations pinned to the version. Authoring happens on mobile/tablet; the web shows them read-only.
Free, up to 5 members per team. Need more seats? Scale out with On-Premise.
Work reports hold your company's code context and decisions — in-house assets. Team collaboration is free and hosted for up to 5 members, but when you need more seats or your data must stay behind your firewall, that's On-Premise. Keep reports inside your own infrastructure and add as many members as you need.
Pricing, tiers, integration, security — the questions people ask most before trying md-log.
Yes. Individual use is a free, operator-hosted SaaS with no credit card and no sales call. Team collaboration is also free in the same SaaS for up to 5 members per team. When you need more members or want data kept in-house, you scale up to self-hosted On-Premise.
Three tracks: Individual is a free single-person SaaS; Team collaboration is free for up to 5 members who share documents and folders to review together; On-Premise is self-hosted with no member limit, keeping reports on your own infrastructure behind your firewall.
Claude Code, Claude Desktop, OpenAI Codex, Cursor, and any other MCP client. Connect by one URL (https://mcp.md-log.com/mcp) with your minted key as a Bearer header — no install, no Node.js.
Yes. Inside the free SaaS you create a team, invite teammates by email (up to 5 per team), and share documents or whole folders to read and annotate together. Split access into Viewer and Editor roles; shared items collect in a ‘Shared with me’ group on web and mobile.
On the operator's servers in the hosted SaaS, or on your own infrastructure behind your firewall with On-Premise. Because work reports carry code context and decisions, On-Premise lets you keep everything inside your firewall when it must not leave for an outside cloud.
Accounts are protected with optional two-factor authentication (2FA), and the MCP key (PAT) you connect to your agent is document-only — a key alone can't manage the account or change settings. Data is protected with encryption at rest and in transit.
No. md-log isn't an automated code reviewer; it's a review viewer and archive where a person reads the report the agent leaves and approves or rejects it. The idea is ‘review the report, not the diff’, and the decision always rests with the human (human-in-the-loop).