Quickstart — Get started in ~5 minutes

md-log is a human-in-the-loop review and archive layer where an AI coding agent writes its work as Markdown (.md) reports and a person reviews them comfortably. This guide gets you from zero to your first reviewed report in about 5 minutes: sign up, connect, save, review.

What you need first

Three prerequisites before you start:

• A free md-log account — individual use is a free, operator-hosted SaaS. No credit card, no sales inquiry.
• Node.js 22 or newer — the MCP connector runs on Node.js 22+. Check with node -v in your terminal. If you don't have it, install the LTS from nodejs.org.
• An AI coding agent — Claude Code, Claude Desktop, OpenAI Codex, Cursor, or any MCP-capable client.

Step 1 — Sign up free

Create your account at app.md-log.com. Individual use is free and needs no credit card.

Step 2 — Mint your MCP key

After signing in, go to Settings → Tokens and mint an MCP key (a personal access token). It looks like mdlog_pat_... and is shown only once at creation, so copy it immediately and keep it somewhere safe.

This key is document-scoped: on its own it can't change your password, mint other keys, or manage login sessions, so a leaked key doesn't hand over your whole account. If you suspect a leak, revoke it in Settings → Tokens and mint a new one.

Step 3 — Connect your agent

The easiest way is to connect by URL — no install, no Node.js. Just the remote endpoint plus the mdlog_pat_... key you copied in Step 2, sent as a Bearer header.

Claude Code (terminal):

claude mcp add --transport http md-log https://mcp.md-log.com/mcp \
  --header "Authorization: Bearer mdlog_pat_YOUR_KEY"

For Cursor, Claude Desktop, and any JSON client, set type to http, url to https://mcp.md-log.com/mcp, and the key as an Authorization: Bearer header. Verify with claude mcp list or /mcp in chat. For per-client setup (remote and local), see the Connect an MCP client guide. The local npx method still works too.

Step 4 — Have your agent save a report

Once connected, your agent can save, edit, and search .md documents by path using md-log's 15 tools. Folders are created automatically on save (like mkdir -p).

To save once, just ask in natural language:

Save this analysis to reports/2026-07-14-first.md.

To avoid asking every time, add an auto-save rule to your project's agent rules file (Claude Code: CLAUDE.md; Codex: AGENTS.md; Cursor: project rules). For example: "When you finish a meaningful task or analysis, write a report summarizing WHAT / WHY / IMPACT and save it via the md-log save_markdown tool at <project>/<YYYY-MM-DD>-<topic>.md, including a commit_message." Then the agent creates a report after each task, images upload too, and every save becomes an immutable version. See the Connect an MCP client guide for details.

Path rules: the filename must end in .md, and .., backslashes, empty path segments, and control characters are not allowed.

Step 5 — Review it

Open app.md-log.com and read the report your agent just saved. It opens the same way in the mobile app or on a tablet. Instead of drowning in the diff, you read the report and approve or reject it yourself — a human-in-the-loop flow.

On pen-capable tablets you can hand-annotate with a stylus (S-Pen on Android, Apple Pencil on iOS; finger-draw fallback where there's no stylus). Each annotation layer pins to a specific document version, and the web renders annotations read-only (an SVG overlay).

Next, read the Review & annotation guide for version history, favorites, and full-text search, and the Connect an MCP client guide for other clients. For team or enterprise on-premise adoption, contact us.