md-feedback MCP server

Read from the package source without running it. The installed server may list more.

• advance_workflow_phase

  Advance workflow phase in strict sequence: scope -> root_cause -> implementation -> verification.

• apply_memo

  Apply an implementation action to a memo. Supports text_replace (requires occurrence or replaceAll when oldText has multiple matches; optional section-scoped propagation), file_patch (applies unified diff patch \u2014 snapshot saved first), and file_create (create a new file). Creates a snapshot before modification, records the implementation, and updates memo status to needs_review.

• approve_checkpoint

  Approve a pending high-risk checkpoint. Grants one execution for the approved tool.

• batch_apply

  Apply multiple implementation operations in a single transaction. Parses the document once, applies all operations sequentially, then writes once. Each operation follows the same format as apply_memo.

• create_annotationwrite action

  Create a new review annotation on a markdown file. Finds the anchor text in the document and attaches a review memo. Auto-creates a quality gate and updates cursor.

• create_checkpointwrite action

  Create a review checkpoint in an annotated markdown file. Records current annotation counts and reviewed sections.

• evaluate_gates

  Evaluate all gates in a markdown file against current memo statuses. Returns updated gate statuses without modifying the file.

• export_review

  Export review feedback in a format optimized for a specific AI coding tool. Targets: claude-code, cursor, codex, copilot, cline, windsurf, roo-code, gemini, antigravity, generic, handoff. Returns formatted markdown ready to save to the appropriate file.

• generate_handoff

  Generate a structured handoff document from an annotated markdown file. Anti-compression format: explicit fields, numbers, lists only.

• get_checkpoints

  List all checkpoints in an annotated markdown file.

• get_document_structure

  Parse an annotated markdown file and return v0.4.0 ReviewDocument metadata. By default bodyMd is omitted to reduce context size; set includeBody=true only when full body is required.

• get_memo_changes

  Get the implementation history and progress for a memo. Returns all MemoImpl records and progress entries from .md-feedback/progress.json. If memoId is omitted, returns all changes.

• get_policy_status

  Return current runtime policy profile and memo-action routing rules.

• get_review_status

  Get current review session status: annotation counts, checkpoints, and reviewed sections. Summary-only \u2014 returns counts and metadata, not individual memos. Use list_annotations for memo details or get_document_structure for the full parse.

• get_severity_status

  Return memo severity overrides and unresolved blocking memo IDs for the document.

• get_workflow_status

  Return current workflow phase and transition history for a document.

• link_artifacts

  Link file artifacts (source files, configs, etc.) to a memo. Creates a MemoArtifact record in the document.

• list_annotations

  List all annotations (USER_MEMO comments) in a markdown file. Returns structured array with id, type, status, owner, text, and color. Lightweight \u2014 returns only memo data, no document body or sections. Use get_document_structure for the full parse.

• list_documents

  List markdown files in the workspace. Optionally filter to only files that already contain annotations.

• pickup_handoff

  Parse an existing handoff document to resume a review session. Returns structured data for session continuity.

• request_approval_checkpointwrite action

  Create an approval checkpoint for a high-risk action before execution.

• respond_to_memo

  Add an AI response to a memo annotation. Inserts a REVIEW_RESPONSE block into the markdown file directly below the memo's anchor text. Automatically sets the memo status to "needs_review" for human approval.

• rollback_memo

  Rollback the latest implementation for a memo. Reverses text_replace operations (swaps before/after), marks the impl as reverted, and sets the memo status back to open.

• set_memo_severity

  Set severity override for a memo. Defaults remain: fix=blocking, question/highlight=non_blocking.

• update_cursorwrite action

  Update the plan cursor position in a markdown file. The cursor tracks "where we are" in a plan. Only one cursor per document.

• update_memo_progresswrite action

  Update the progress of a memo with a status change and message. Writes progress to .md-feedback/progress.json and updates the memo status. Terminal statuses (done, failed) require human approval via VS Code.

• update_memo_statuswrite action

  Update the status of a memo annotation. Writes the change back to the markdown file. Returns the updated memo. Terminal statuses (answered, done, failed, wontfix) require human approval via VS Code.

• Code scan3 source files scanned20/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitylocal package, no credentials required12/15
• Maintenancelast push 188 days ago4/15
• Maintainer identityregistry namespace matches repository owner6/10

Safety score

65/100

higher than 20% of 48,255 servers

Tools

27

the typical server has 7

Tools that change things

22%

typical: 0%. Fewer write tools means less an agent can break

GitHub stars

5

more than 83% of 28,688 servers

Downloads a week

53

more than 44% of 12,191 servers with a package

md-feedback: common questions

Is md-feedback MCP server safe?

With care: it is graded C, so read the findings first (65/100). Read the md-feedback safety report

How do I install md-feedback?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does md-feedback need an API key?

Not as far as the registry entry and our scan can tell: no credentials are declared or required.

Is md-feedback maintained?

The last commit was 226 days ago (2026-03-23). The latest release is v1.5.6.

How does md-feedback compare with other MCP servers?

safety score 65/100 (higher than 20% of 48,255 servers); tools 27 (the typical server has 7); tools that change things 22% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.