How sign-in connections work.
Some clients connect by signing in to ScoutRival instead of pasting a token. Here's what that flow does, what you approve, and how to manage or revoke it.
Sign-in — OAuth — is how claude.ai on the web and ChatGPT connect to ScoutRival. Instead of pasting a token, you sign in to your ScoutRival account and approve the connection. The client then holds a scoped connection to your account, and you're back in the assistant with the same ScoutRival tools a token client gets.
Sign-in is the simplest and safest way to connect a web client. There's no token to copy, store or lose, and you approve exactly what the connection can do on a consent screen before anything is granted.
What sign-in is
Sign-in connections are for clients that run in the browser and don't want you handling a secret string. Claude.ai on the web and ChatGPT use this method: you point the client at ScoutRival, sign in to your account, and approve the connection. Token clients — Cursor, Claude Desktop and Claude Code — paste a connection token instead. The route in is different, but the end result is identical: the same ScoutRival tools, reading the same data, spending credits the same way.
What happens, step by step
Add the server URL
In the client, add ScoutRival's server: https://user.scoutrival.com/api/mcp.
The client opens ScoutRival's sign-in
Instead of asking for a token, the client opens ScoutRival's sign-in page in your browser.
You sign in
Sign in to your ScoutRival account. If you're already signed in, this is a single click.
You choose an access level
A consent screen shows what access is being granted and which brands are in reach. You pick an access level — Read-only, Content or Full.
You Allow, and you're back in the assistant
Select Allow. The client receives a scoped connection and returns you to the chat, ready to use the tools.
No password or token is ever shared with the assistant. The client receives only a scoped connection — it cannot read your password, and it can only do what the access level you chose permits.
Sign-in vs token
Both methods reach the same tools, and both let you choose Read-only, Content or Full. They differ only in how you connect and where the connection lives.
| Sign-in (OAuth) | Token | |
|---|---|---|
| Setup | Click Allow | Paste a string |
| Where it lives | Your client's connection store | A config file you protect |
| Clients | claude.ai web, ChatGPT | Cursor, Claude Desktop, Claude Code, multi-model clients |
| Revoke | Remove the connector in the client's settings | Revoke at Settings → MCP |
The consent screen
The consent screen is where you choose the access level and confirm the brands the connection can reach. Start with Read-only — it reads everything and spends nothing, which is enough to try the connection. When you want the assistant to generate content or run a check, remove the connection and reconnect at a higher level. See scopes and access levels for exactly what each level unlocks.
Manage or revoke
To revoke a sign-in connection, remove it in the client's connector settings. That severs the access immediately. Because nothing sensitive is stored in your chat — no token, no password — removing the connector is all that's needed to cut it off.
If a connection stops working or looks stale, remove it and add it again. Reconnecting also refreshes the tool list, which fixes a client that has cached an old set of tools.