Review, annotate & version your reports — the reviewer's guide

md-log turns your AI coding agent's output into something a human can actually review. Instead of scrolling a giant diff, you read a Markdown report the agent wrote — what it did, why, and the impact — then approve or reject it. This guide covers where reports appear, how to review them on any device, how stylus annotation works, and how immutable versions preserve the whole history.

Where your reports show up

Your agent writes reports by file path into md-log over MCP (see Connect MCP). Reports live in a hierarchical folder tree, so a path like my-project/2026-07-14-auth-refactor.md places the report exactly where you expect. Folders auto-create on save, the way mkdir -p does — the agent never has to create them first. Images and other assets upload alongside the text.

Once a report is saved you can find it three ways:

• Browse the folder tree on the web app at app.md-log.com or in the mobile app.
• Full-text search across your reports.
• Mark reports you return to often as favorites.

For the bigger picture of how the pieces fit together, see the Overview.

Reading & reviewing — human-in-the-loop

The review model is deliberately simple: the surface is the report, not the diff. You read the agent's summary of what changed and why, and you make the call — approve or reject. md-log is a review viewer and archive; it does not auto-review code or replace your judgment. You are the reviewer.

Because reports render cleanly on web, phone, and tablet, review is asynchronous — you can vet an agent's work from anywhere, away from your desk, and catch up on what your agents did while you were out. New teammates (and new agents) can read back through the accumulated reports as onboarding material.

Stylus annotation on pen-capable tablets

On tablets with a pen, you can hand-annotate a report directly on top of the rendered Markdown:

• S-Pen on Android tablets (native).
• Apple Pencil on iOS (iPad).
• Finger-draw fallback on devices without a stylus.

Pen-only controls appear only on pen-capable devices, so the UI stays clean where there's no stylus. Stroke authoring happens on mobile and tablet — that's where you draw.

Annotations are version-pinned

Each annotation layer belongs to one specific document version. md-log does not silently re-anchor or carry your ink forward when the document changes — this keeps every annotation truthful to the exact version you were looking at when you drew it. If you want the same marks on a newer version, use the explicit copy action to duplicate them verbatim onto that version.

If the text block a stroke was drawn on no longer exists in the version you're viewing, those orphaned strokes park at the bottom (handled client-side) rather than floating over unrelated content.

The web app is read-only for annotations

The web app renders annotations but never authors them. On the web your ink appears as an SVG overlay pinned to the annotated version, so reviewers on a laptop can see exactly what was marked up — but to draw new strokes you use the mobile or tablet app. This split keeps web review lightweight and precise.

Immutable versions & history

Every save is a new immutable version. Nothing is overwritten in place, so your project's AI-collaboration history accumulates over time.

• Every save/update/append carries a commit message that shows up in the version history.
• You can view the full history and open and read any past version.
• Move or rename a report and its history and annotations are preserved.

When two edits collide, md-log uses optimistic concurrency: a conflicting write returns a sync conflict along with the server's current head, so nothing is clobbered by accident. A force option is available for last-writer-wins when you deliberately want it.

Offline-first mobile, search & favorites

The mobile app (iOS + Android, phones & tablets) is offline-first — it keeps a local database and syncs deltas when you're back online, so you can read and annotate on the move. Across every surface you get full-text search and favorites to get back to the reports that matter.

Ready to wire up your agent? Start with Connect MCP, or read the Overview first.