Does Appbot, AppFollow, or Appfigures Have an MCP Server? The 2026 Matrix

Short answer, as of July 2026: Appbot ships no MCP server we can find, official or community. AppFollow has no official one either — only third-party wrappers (Zapier, MCPBundles) that expose a handful of read-only tools. Appfigures is the one with a genuine official MCP, marketed as "Appfigures for AI agents," but it's built around reading analytics and listing reviews, not posting replies. So if your question is "can I run my whole review queue from Claude with one of these," the honest answer is: not the way you probably mean.

That distinction, read-only versus a connector that can actually act, is the whole game, and it's what this matrix is about. Every claim below is dated. Appbot, AppFollow, and Appfigures are trademarks of their respective owners; ReplyArgus is not affiliated with or endorsed by any of them. We build a competing tool, so read this as a comparison from an interested party who checked the docs before writing.

Here's where each tool stands on shipping an MCP server and what that server can do, as of July 2026. "Reads" means it can pull data into an assistant; "acts" means the connector can draft and move a reply toward being published.

No, not that we could find as of July 2026, from Appbot directly or from the community. Appbot is a review-monitoring and sentiment-analysis platform: it tags themes, tracks sentiment over time, and slices reviews across App Store and Google Play. But that work happens inside Appbot's own dashboard. There's no exposed set of MCP tools for an assistant like Claude or ChatGPT to call, so you can't say "triage my 1-star reviews and draft replies" from a chat and have it reach Appbot's data.

That may change; plenty of SaaS tools are adding connectors this year. But a roadmap isn't a shipped feature, and if operating from your AI tools matters to you today, Appbot isn't the one that does it yet. If you're weighing the platform overall, we break down where it fits in our [Appbot comparison](/vs/appbot).

Not an official one. What exists is third-party: AppFollow's REST API has been wrapped into MCP servers by platforms like Zapier and MCPBundles. Those wrappers surface roughly five read-only tools — list your app collections, list tracked apps, pull ratings, pull review stats, list users. Useful for asking an assistant "what's my average rating this week," but strictly reporting. There's no tool in the set that drafts a reply, let alone posts one.

Two things to keep in mind about a third-party wrapper. First, it's read-only by design, so it can't close the loop from complaint to posted reply. Second, you're routing your AppFollow API key through a middleman platform, which is a trust and auth decision worth making deliberately rather than by default. If you're weighing AppFollow overall, we go deeper in our [AppFollow comparison](/vs/appfollow).

Yes, and it's the most capable of the three. Appfigures ships an official MCP server ("Appfigures for AI agents") that exposes nineteen tools to any MCP client: Claude Desktop, Cursor, and the rest. It's a real, first-party connector, which is more than either of its two rivals here can say.

But look at what those nineteen tools do. Every one is a read: list products, list reviews, get the ratings report, revenue report, subscriptions report, ad-spend report, and so on. One of the example prompts Appfigures itself markets is "pull this week's 1-star reviews, grouped by app version, and draft replies I can approve" — and the assistant can draft in the chat window, but the connector's job stops at pulling data. The actual posting of a reply lives in Appfigures' dashboard flow and requires a linked store account on your plan. On the MCPBundles listing the server is described outright as read-only, and answers the publish question directly: agents can read analytics, reports, and reviews, but the server "does not publish store listings or change account settings." So even the best competitor MCP here is a reporting-and-drafting surface, not an operate-your-queue-end-to-end one.

Replying is where ratings actually move. Google's I/O 2019 data showed apps average +0.7 stars after developers start responding, and Hassan et al.'s study of 4.5M reviews found users were roughly 6× more likely to raise their rating after a reply (4.4% vs 0.7%). A read-only MCP surfaces the reviews that need that response; it doesn't write or post the response. You're still doing the slow half by hand.

This is the gap [ReplyArgus](/features) is built to close, and it's why its MCP connector is designed to act, not just report. From inside Claude, ChatGPT, or Cursor, you can ask it to draft replies to every unanswered review — each grounded in your own past approved replies and knowledge base, each written in the reviewer's own language across 100+ languages, each landing in an approval queue rather than posting silently. And unlike an analytics tool, it publishes: to the App Store and Google Play both, either on your approval or via opt-in rules you set by rating, keyword, or language. Here's the kind of draft it hands back for a furious one-star:

That reply is grounded, not improvised — it names the real bug, offers a next step, and promises no date it can't keep. Getting the language right is its own discipline; we cover it in [replying to app reviews in any language](/blog/reply-to-app-reviews-in-any-language). One mechanical note the tools that can post have to respect: Google Play caps replies at 350 characters, hard. Apple publishes no official limit, and community testing suggests a few thousand characters, so it's safest to write to the tighter Play ceiling. A read-only MCP never has to think about that, because it never posts.

If you want the full walkthrough of running a review queue from a chat — the exact prompts, the OAuth setup, the guardrails — we wrote it up in [managing app reviews from Claude](/blog/manage-app-reviews-from-claude).

If you're comparing connectors, don't stop at "does it have an MCP." Ask what the MCP can actually do: