# ReviewResponder

> Draft replies to your reviews

ReviewResponder helps restaurant, salon, dental, auto repair, hotel, gym, retail, home-service, clinic and cafe owners answer the Google and Yelp reviews they paste in: draft one reply in a warm, professional or concise tone that thanks, addresses specifics, apologizes without arguing and takes negatives offline; draft a prioritized batch of replies (1-2 stars first, oldest first); find recurring praise and complaint themes with sample quotes and suggested fixes; estimate the hours per month and rating lift of answering consistently with built-in typical estimates by business type; build a response-time policy with a weekly schedule, escalation triggers and a template bank; and scan a drafted reply for guideline or legal red flags such as incentives, rating-change requests, private or health details, threats and admissions of liability. Drafts and analysis only - it never fetches reviews from any platform, never posts anything, and gives no legal advice. You edit and post from your own accounts.

## Use from an AI assistant

- MCP endpoint (Streamable HTTP, JSON profile): POST https://review-responder.magicteams.ai/mcp
- Directories: ChatGPT Apps, Claude Connectors, Meta Muse (search "ReviewResponder")
- Machine manifest: https://review-responder.magicteams.ai/.well-known/agent.json
- UCP discovery profile: https://review-responder.magicteams.ai/.well-known/ucp

## Tools

- draft_review_reply: Draft review reply — Draft one owner reply to a Google or Yelp review you paste in, from the review text, star rating, business name and tone (warm, professional or concise), following platform best practice: thank, address the specifics named, apologize for negatives without arguing, move 1-3 star issues offline, under 120 words, no incentives. Returns the draft plus a checklist. Use for a single review. Do NOT use for several reviews at once; use batch_draft_replies.
- batch_draft_replies: Batch draft replies — Draft replies for 1-25 pasted reviews and order them by urgency: unreplied 1-2 star reviews first, then 3 star, then 4-5 star, oldest first within each band. Returns a numbered reply order with a draft, urgency and reason for each, plus counts. Use when catching up on a backlog. Do NOT use to find recurring themes; use analyze_review_themes.
- analyze_review_themes: Analyze review themes — Summarize 1-50 pasted reviews: rating distribution and average, recurring positive and negative themes from a fixed keyword lexicon (service, wait time, price, cleanliness, staff, quality, booking, parking, communication, atmosphere, accuracy) with mention counts and a sample quote each, top praise and complaint quotes, and a suggested fix per negative theme. Use to see what customers keep saying. Do NOT use to write replies; use batch_draft_replies.
- estimate_response_impact: Estimate response impact — Estimate what answering reviews consistently is worth for a local business: replies needed per month at a target response rate (default 90%), hours per month with drafts versus writing by hand, a rating-lift range over 12 months and revenue at stake from average ticket and monthly customers. Built-in typical estimates per business type (restaurant, salon, dental, auto_repair, hotel, gym, retail, home_services, clinic, cafe) fill any number you do not give; ranges are labelled planning assumptions. Use to decide whether a reply routine is worth the time. Do NOT use to set response-time targets; use build_response_sla.
- build_response_sla: Build response policy — Build a review-response policy for your team: target response time per star band (default 1-2 stars within 24h, 3 stars 48h, 4-5 stars one week), who drafts and who approves, a weekly check schedule sized to your review volume, escalation triggers and a template bank per star band with placeholders. Use when setting up a routine. Do NOT use to size the benefit; use estimate_response_impact.
- flag_policy_risks: Flag policy risks — Scan a drafted reply (or a review) for wording that breaks Google or Yelp guidelines or creates legal exposure: offering incentives, asking for a rating change, soliciting fake reviews, revealing private or health details, legal threats, admitting liability, discriminatory language, arguing, profanity. Returns flags with severity, the matched text, why it matters and a fix, plus a risk level. Use before posting any reply. Do NOT use to write the reply; use draft_review_reply.

All tools are free, compute per request, and store nothing.

## Operator

MagicTeams — support@magicteams.ai — https://review-responder.magicteams.ai/support