• Canvas MCP server must be running and connected to the agent's MCP client (e.g., Claude Code, Cursor, Codex, OpenCode).
• The authenticated user must have an educator or instructor role in the target Canvas course.
• The assignment must have peer reviews enabled in Canvas (either manual or automatic assignment).
• FERPA compliance: Set

  ENABLE_DATA_ANONYMIZATION=true

  in the Canvas MCP server environment to anonymize student names. When enabled, names render as

  Student_xxxxxxxx

  hashes while preserving functional user IDs for messaging.

Which course and assignment would you like to check peer reviews for?

list_courses

list_assignments

get_peer_review_completion_analytics

• Overall completion rate (percentage)
• Number of students with all reviews complete, partial, and none complete
• Per-student breakdown showing completed vs. assigned reviews

get_peer_review_assignments

• include_names=true

  for human-readable output

• include_submission_details=true

  for submission context

get_peer_review_comments

• include_reviewer_info=true

  -- who wrote the comment

• include_reviewee_info=true

  -- who received the comment

• anonymize_students=true

  -- recommended when sharing results or working with sensitive data

analyze_peer_review_quality

• Average quality score (1-5 scale)
• Word count statistics (mean, median, range)
• Constructiveness analysis (constructive feedback vs. generic comments vs. specific suggestions)
• Sentiment distribution (positive, neutral, negative)
• Flagged reviews that fall below quality thresholds

analysis_criteria

identify_problematic_peer_reviews

• Very short or empty comments
• Generic responses (e.g., "looks good", "nice work")
• Lack of constructive feedback
• Potential copy-paste or identical reviews

criteria

get_peer_review_followup_list

• priority_filter="urgent"

  -- students with zero reviews completed

• priority_filter="medium"

  -- students with partial completion

• priority_filter="all"

  -- everyone who needs follow-up

• days_threshold=3

  -- adjusts urgency calculation based on days since assignment

send_peer_review_reminders

• recipient_ids

  -- list of Canvas user IDs from the analytics results

• custom_message

  -- optional custom text (a default template is used if omitted)

• subject_prefix

  -- defaults to "Peer Review Reminder"

1. Get incomplete reviewers from step 2
2. Extract their user IDs
3. Review the recipient list with the user
4. Send reminders after confirmation

send_peer_review_followup_campaign

1. Runs completion analytics automatically
2. Segments students into "urgent" (none complete) and "partial" groups
3. Sends appropriately toned reminders to each group
4. Returns combined analytics and messaging results

extract_peer_review_dataset

• output_format="csv"

  or

  output_format="json"

• include_analytics=true

  -- appends quality metrics to the export

• anonymize_data=true

  -- recommended for sharing or archival

• save_locally=true

  -- saves to a local file; set to

  false

  to return data inline

generate_peer_review_feedback_report

• report_type="comprehensive"

  -- full analysis with samples of low-quality reviews

• report_type="summary"

  -- executive overview only

• report_type="individual"

  -- per-student breakdown

• include_student_names=false

  -- recommended for FERPA compliance

generate_peer_review_report

save_to_file=true

• Completion rate: 72% (23/32 students fully complete)
• Partial: 5 students (started but not finished)
• Not started: 4 students

• Student_a8f7e23 (0 of 3 reviews done)
• Student_b2c91d4 (0 of 3 reviews done)
• Student_f5e67a1 (0 of 3 reviews done)
• Student_d9c34b2 (0 of 3 reviews done)

• Student_c1d82e5 (1 of 3 reviews done)
• Student_e4f03a9 (2 of 3 reviews done)

**User:** "Send reminders to the ones who haven't started"

**Agent:** Confirms the 4 recipients, then calls `send_peer_review_reminders` with their user IDs.

**User:** "Now check if the completed reviews are any good"

**Agent:** Calls `analyze_peer_review_quality` and presents quality scores, flags 3 reviews as too short, and recommends the instructor follow up with specific students.

• Confirm before sending -- Always present the recipient list and message content to the instructor before calling any messaging tool.
• Use dry runs -- When testing workflows, start with a single recipient or confirm the output of analytics tools before acting on the data.
• Anonymize by default -- Use

  anonymize_students=true

  or

  anonymize_data=true

  when reviewing data in shared contexts.
• Respect rate limits -- The Canvas API allows roughly 700 requests per 10 minutes. For large courses, the messaging tools send messages sequentially with built-in delays.
• FERPA compliance -- Never display student names in logs, shared screens, or exported files unless the instructor has explicitly confirmed the context is appropriate.

• Peer reviews must be enabled on the assignment in Canvas before any of these tools return data.
• The

  send_peer_review_followup_campaign

  tool combines analytics and messaging into one call -- powerful but sends real messages. Use it only after confirming intent with the instructor.
• Quality analysis uses heuristics (word count, keyword matching, sentiment). It identifies likely low-quality reviews but is not a substitute for instructor judgment.
• This skill pairs well with

  canvas-morning-check

  for a full course health overview that includes peer review status alongside submission rates and grade distribution.