App Reviews in Claude Code: Setting Up the ReplyArgus MCP Server
Connect the ReplyArgus MCP to Claude Code, then run real prompts: summarize this week's complaints, draft a grounded reply, triage a bug — without leaving the terminal.
The Argus Team
Reply Argus
You can run your entire App Store and Google Play review queue from inside Claude Code — no dashboard tab, no copy-paste. Add the ReplyArgus MCP server once, and Claude can list this morning's 1-star reviews, cluster the complaints, draft a reply grounded in your own past answers, and (if you let it) publish. The whole thing is one `claude mcp add` command and an OAuth login.
This is a setup guide. First the connection, then three prompts you'll actually reuse: *summarize this week's complaints*, *draft a grounded reply*, and *turn a review into a bug ticket*. Everything below uses the real tools the server exposes — nothing hand-wavy.
What the ReplyArgus MCP server exposes
The [Model Context Protocol](/blog/app-reviews-mcp) is just a standard way for an AI client to call your tools. ReplyArgus ships a hosted MCP server so any MCP-aware client (Claude Code, the Claude desktop app, ChatGPT, Cursor) can drive your review inbox with your permissions, not a scraper's. Roughly, the tools fall into four buckets:
- Read — `list_apps`, `list_reviews`, `get_review`, `search_reviews`, and `sync_reviews` to pull the latest from Apple and Google.
- Understand — `theme_breakdown`, `weekly_report`, and `analytics_overview` cluster reviews into recurring themes and trends instead of a raw dump.
- Write — `draft_reply` (grounded in your approved replies + knowledge base), `translate_review`, and `publish_reply` when a human signs off.
- Operate — `set_auto_publish`, `set_draft_cadence`, `toggle_agent`, plus `usage_and_limits` and `whoami` so Claude always knows which workspace and quota it's touching.
One connector, every client
The same server works in Claude Code, Claude desktop, Cursor, and ChatGPT — the setup differs only in how each client registers a remote MCP endpoint. This guide covers Claude Code (the CLI); the [manage-app-reviews-from-Claude walkthrough](/blog/manage-app-reviews-from-claude) covers the desktop app.
Connect it to Claude Code
ReplyArgus runs a remote Streamable-HTTP server at `www.replyargus.com/api/mcp`, authenticated with OAuth — so you never paste a raw API key into a config file. You approve access in a browser, and Claude Code holds the token.
- 1
Step 1 — Add the server
In your terminal, run `claude mcp add --transport http replyargus www.replyargus.com/api/mcp`. That registers the remote endpoint under the name `replyargus`.
- 2
Step 2 — Authenticate
Inside Claude Code, run the `/mcp` command. It lists your servers and shows `replyargus` as needing auth. Trigger the login and a browser opens on the ReplyArgus OAuth screen — sign in and approve. The token is stored for you.
- 3
Step 3 — Verify
Ask Claude: “Using ReplyArgus, run whoami and list my apps.” If it returns your org and app names, you're connected. If it 401s, your token expired — re-run `/mcp` to re-auth.
- 4
Step 4 — Sync
Ask it to `sync_reviews` for an app before your first real prompt, so Claude is working from the freshest pull rather than yesterday's cache.
The connector is Owner/Admin only
Only Owners and Admins can authorize the MCP connector — it's gated at a single server chokepoint, so a downgraded teammate fails closed on their next call. If the OAuth step succeeds but every tool returns a permission error, check your role on the [pricing and roles page](/pricing). Members, drafters, and viewers work in the app UI, not through the connector.
Prompt 1 — Summarize this week's complaints
The dashboard shows you a list. Claude Code can show you the *pattern*. A prompt like this leans on `weekly_report` and `theme_breakdown`:
> “Using ReplyArgus, pull this week's reviews for Acme Notes across both stores, group the 1- and 2-star ones by theme, and tell me the top three complaints with a representative quote for each.”
You get back something you can paste into a standup: *sync failures on iOS 18 (11 reviews), a paywall that fires on the free tier (7), and export-to-PDF crashes (5)*. Because it's reading structured review data, not screenshots, the counts are real, and you can immediately drill in: “Show me the full text of the three worst sync-failure reviews.”
Ask for `analytics_overview` alongside the summary and Claude will tell you whether that theme is rising or fading week over week. Trends matter because [recent reviews weigh more](/blog/recent-reviews-weigh-more) in your store rating than old ones — a spike this week hurts more than the same count six months back.
Prompt 2 — Draft a grounded reply
This is where an MCP server beats pasting a review into a chat window. `draft_reply` doesn't freestyle — it grounds every reply in your previously approved replies plus your knowledge base (store listing + marketing page, auto-ingested), so the tone matches your brand and the facts match your product. The difference between [grounded and hallucinated replies](/blog/grounded-vs-hallucinated-ai-replies) is the difference between “we fixed this in 4.2, update from the App Store” and inventing a version number that never shipped.
Try: “Using ReplyArgus, draft a reply to the top sync-failure review. Keep it under 350 characters so it fits Google Play, apologize once, and point them to the 4.2 update.” Claude calls the tool and hands you a draft you can edit in place before anything is published.
Was great until the last update — now my notes won't sync between my phone and iPad. Lost a whole meeting's worth of notes. Fix this.
We're sorry — losing notes is exactly what sync is supposed to prevent, and we take that seriously. The 4.2 update (live now in both stores) fixes the cross-device sync bug. Update, then reopen the affected note to force a re-sync. If anything's still missing, email support@acme.app and we'll recover it.
Note what Claude did *not* do: publish it. ReplyArgus is approve-by-default. The draft sits until a human (you, right there in the terminal) says ship it. That's the safe posture, and it's why teams trust the connector to touch a public-facing surface. If you want hands-off publishing for the easy cases, that's a deliberate, [narrow auto-publish rule](/blog/is-it-safe-to-auto-publish-app-review-replies) you opt into, not a default.
Prompt 3 — Turn a review into a bug ticket
Reviews are your cheapest bug tracker, but only if the signal gets to engineering. Claude Code can bridge that gap. Use `search_reviews` to gather every instance of a bug, then ask Claude to write it up:
> “Using ReplyArgus, search all Acme Notes reviews for ‘sync’ from the last 30 days, count how many mention data loss, and draft a GitHub issue with a title, repro steps inferred from the reviews, affected OS versions, and three verbatim quotes as evidence.”
You get a ready-to-file issue backed by real user reports. ReplyArgus already clusters reviews into a PM roadmap board that exports to Jira, Notion, Google Sheets, and DevRev — so the connector is the fast path (draft the ticket from the terminal now) and the board is the durable path (a running backlog your PM owns). Use whichever fits the moment.
Why do this in Claude Code at all?
Because your reviews now live in the same place as your code, your terminal, and your other MCP tools. You can go from “what are people complaining about” to a drafted reply to a filed issue in one thread, without context-switching to a browser. And because Claude reads structured data through the connector, the answers are grounded in fact rather than a scrape of a screenshot.
It also composes. Point Claude Code at your repo and the ReplyArgus MCP in the same session, and you can ask it to cross-reference a crash in your logs against the reviews reporting it — something no dashboard does. The [full agentic-tools overview](/agentic-tools) covers the other clients and the complete tool list.
Frequently asked
- What's the ReplyArgus MCP endpoint for Claude Code?
- It's `www.replyargus.com/api/mcp`, a remote Streamable-HTTP server. Add it with `claude mcp add --transport http replyargus www.replyargus.com/api/mcp`, then run `/mcp` inside Claude Code to complete the OAuth login.
- Do I need an API key?
- No. The connector uses OAuth, so you approve access in a browser and Claude Code stores the token. You never paste a raw key into a config file, and access follows your ReplyArgus role and permissions.
- Can the MCP server publish replies automatically?
- Only if you tell it to. ReplyArgus is approve-by-default: `draft_reply` produces a draft that waits for a human, and `publish_reply` runs only when you approve. Rule-based auto-publish is a separate, opt-in setting scoped by rating, keyword, language, or store.
- Who on my team can connect the MCP server?
- Owners and Admins only. The connector is gated at a single server checkpoint, so members, drafters, and viewers can't authorize it — they work in the app UI instead. A demoted teammate loses connector access on their next call.
- Which stores does it cover?
- Apple App Store and Google Play. The server reads and drafts across both in one place; the [App Store vs Google Play reply mechanics differ](/blog/app-store-vs-google-play-review-replies) (Google enforces a hard 350-character reply limit; Apple publishes no official cap), and ReplyArgus handles each store's rules for you.
- Does it work in ChatGPT or Cursor too?
- Yes. The same server is MCP-standard, so any MCP-aware client can register the endpoint. Only the way you add a remote connector differs per client.
Wire it up and your review queue becomes just another thing Claude Code can do. Start free — add the ReplyArgus MCP, run `whoami`, and have Claude draft your first grounded reply in the next five minutes. [Create your workspace](/signup) (no card for the Free tier), then paste in the `claude mcp add` command above.
Try it
Let Argus draft your next reply.
Watch it answer a real review in your voice. 10-day trial, no card to begin.
Keep reading
Manage Your App Store Reviews From Claude: The 5-Minute MCP Setup
Connect one MCP server and run your App Store + Google Play review queue from a Claude chat: summarize complaints, draft grounded replies, queue for approval.
Read moreThe MCP Server for App Reviews: Connect Your AI Agent to Your Store Reviews
An app reviews MCP server hands your live App Store and Google Play reviews to Claude, ChatGPT, or Cursor as callable tools. Here's what it exposes and how to wire it up.
Read more