Quick Start
Get up and running with MRSF in under 2 minutes.
This is the fastest way to start adding Markdown comments in a durable way, without embedding brittle comments directly in Markdown.
Install the CLI
npm install -g @mrsf/clipip install mrsfOr use without installing:
npx @mrsf/cli --helpCreate a Sidecar
mrsf init docs/architecture.mdThis creates docs/architecture.md.review.yaml with the basic structure.
Add a Comment
mrsf add docs/architecture.md -l 12 "This section needs more detail."This adds an anchored comment at line 12, automatically capturing the selected_text and current git commit.
You can attach tool-specific metadata with repeatable --ext key=value flags. These values are persisted as flat x_* fields on the comment:
mrsf add docs/architecture.md \
--author "review-bot" \
--text "Needs a second pass" \
--line 12 \
--ext x_source=review-bot \
--ext x_score=0.91 \
--ext 'x_labels=["needs-review","docs"]'Validate
mrsf validateDiscovers all sidecars in the workspace and checks them against the MRSF schema.
Re-anchor After Edits
After editing the Markdown document:
mrsf reanchorThis finds where anchored text has moved and updates line numbers. Use --dry-run to preview changes.
Check Status
mrsf statusReports anchor health for every comment:
| Health | Meaning |
|---|---|
| fresh | Text matches at the recorded position |
| stale | Commit is behind HEAD; text may have moved |
| orphaned | Text can't be found in the current document |
List & Filter Comments
# All open comments
mrsf list --open
# Summary statistics
mrsf list --summary
# Filter by author
mrsf list --author "Jane Doe"Next Steps
- Examples — worked examples for every re-anchoring strategy
- Full CLI Reference — all commands and options
- Python CLI & SDK — use MRSF from Python
- MCP Server — expose MRSF to AI assistants
- Specification — the complete MRSF v1.0 spec
