MCP SETUP GUIDE
Connect AppGazers to Claude Code and Claude for App Research
Use AppGazers with Claude Code remote HTTP MCP or a hosted Claude custom connector, with OAuth, source checks, and clear surface-specific limits.
Claude Code and hosted Claude both support MCP in documented but different ways. Claude Code can add AppGazers as a remote HTTP server. Hosted Claude can use a custom remote connector where the plan and organization settings allow it. In either case, use https://appgazers.com/mcp and complete OAuth in the browser.
At a glance
| Claude surface | Documented route | Boundary |
|---|---|---|
| Claude Code | Remote HTTP MCP command, then browser login | A local development client, not hosted Claude configuration |
| Hosted Claude | Customize → Connectors → custom remote connector where available | Plan and organization policy may control access |
| AppGazers access | OAuth to https://appgazers.com/mcp | mcp:user is not needed for read-only research |
Connect from Claude Code
In a terminal, add the remote HTTP server and start its OAuth login. Claude Code’s MCP reference documents remote HTTP support and the CLI manages the browser authorization flow.
Try this prompt
Terminal commands:
claude mcp add --transport http appgazers https://appgazers.com/mcp
claude mcp login appgazers
Assistant prompt:
Use AppGazers to identify three competitors for [app idea] in [country]. Cite store links, label estimates, and state uncertainty.Connect from hosted Claude only where available
Anthropic’s custom-connector guide describes the hosted path: Customize → Connectors → add a custom connector, paste the public URL, then connect and authorize. Organization owners may add a connector for members; plan and admin rules apply.
This is not a substitute for the Claude Code command. If the connector option is absent, verify plan eligibility and organization controls with Anthropic instead of sharing credentials or trying an unsupported transport.
Use a bounded prompt and inspect the tool trail
Ask Claude to retrieve evidence for a small app set, not to announce an unverified market winner. Require storefront, date, source links, and a distinction between observed records, third-party estimates, and its own synthesis.
Try this prompt
Using AppGazers MCP, build a decision memo for [category] in [country]. Retrieve no more than five apps. Include source links, current catalog/ranking evidence, review themes, keyword context, and public-ad evidence if present. Put estimates in a separate column. Then list assumptions and what a human should verify before acting.Handle failures and protect the account boundary
If Claude Code lists the server as needing authentication, run the login command again and finish the browser flow. If a hosted connector cannot reach the endpoint, check that it is configured as public remote MCP and that the exact HTTPS URL is used. Never replace OAuth with a copied secret.
Approve only scopes you recognize. mcp:read enables catalog research; mcp:user is for watchlist changes. Revoke unused connections from the AppGazers MCP page and check every result against its cited source before using it in a roadmap or campaign.
Official sources reviewed
- Anthropic: Claude Code MCP — Official Claude Code transport and command reference.
- Anthropic: custom remote connectors — Official hosted Claude connector guidance.
- AppGazers MCP — Endpoint and OAuth scope guidance.
RESEARCH YOUR NEXT APP
Start with a niche. Leave with evidence.
AppGazers is free during beta. The research workspace opens after Google sign-in.
Explore rising apps