Connect Claude Desktop and Claude Web
Add RankX AI to Claude as a custom connector: open Customize, then Connectors, paste the MCP server URL, let Claude register itself and sign in with OAuth.
Claude connects to RankX AI as a custom connector, signed in with OAuth. Go
to Customize, then Connectors, add a custom connector with the address
https://app.rankxai.com/api/mcp, let Claude register itself, and sign in. An
account owner approves the connection once and it stays connected.
A custom connector belongs to your Claude account, so the same connection works on claude.ai, in Claude Desktop and on mobile. Do not set it up by editing Claude Desktop's config file: that file runs local programs only, and the failure is silent.
How do I add RankX AI as a custom connector?
On a Free, Pro or Max plan you add the connector to your own Claude account. Claude's Free plan allows one custom connector.
Open Connectors
In Claude, go to Customize, then Connectors, and select Add custom connector. Some accounts show a + Add button first.
Name it and paste the endpoint
Name the connector RankX AI, then enter the MCP server URL:
https://app.rankxai.com/api/mcpLet Claude register itself
The dialog differs between accounts, and both versions work with RankX AI:
| If the dialog shows | Choose |
|---|---|
| Authentication and OAuth client choices | Sign in now, then Register automatically |
| One screen with Advanced settings | Leave the client ID and secret empty |
RankX AI supports dynamic client registration, so Claude registers itself and needs no client ID or secret. An invented client ID makes the sign-in fail. Select Add.
Sign in and approve
Claude sends you to RankX AI to sign in. The approval screen says Claude will be able to read your data, run actions that use credits and publish changes to your connected websites. See permissions lists all six capabilities and marks the three that change a live site.
Only an account owner can approve. Anyone else sees a message asking the owner to connect the app, and nothing is granted. Select Approve connection.
Check it worked
In a chat, select +, then Connectors, and make sure RankX AI is on. Then ask:
Using RankX AI, list my websites.
Claude should call list_projects and return them. If no RankX AI tools appear,
see troubleshooting.
How does the Claude sign-in flow work?
Claude signs in to the RankX AI MCP server with OAuth, and every step after your approval happens in the background. RankX AI supports each part of the flow Anthropic documents for custom connectors.
- Claude calls
https://app.rankxai.com/api/mcpfrom Anthropic's servers. RankX AI answers401with a pointer to its OAuth metadata. - Claude reads the metadata, registers itself as a client, and opens RankX AI's approval screen for you.
- After you approve, Claude receives an access token that lasts an hour and a refresh token it uses to keep access current.
The user who approves sees every capability before it is granted, and on an account with several clients chooses which websites the connection can reach. Claude then lists the RankX AI tools in the chat; each tool states what it does and whether it uses credits.
How do Team and Enterprise plans add it?
On Claude Team and Enterprise, an organisation Owner adds the connector once and each member then connects with their own sign-in.
- The Owner goes to Organization settings, then Connectors, selects Add, then Custom, and chooses Web if asked.
- The Owner enters
https://app.rankxai.com/api/mcpand selects Add, leaving the OAuth client ID and secret empty. - Each member goes to Customize, then Connectors, finds RankX AI with the Custom label and selects Connect.
Each member signs in to RankX AI separately. Only a RankX AI account owner can approve a connection, so the member who connects should be the owner of the RankX AI account it reaches.
Why does the Claude Desktop config file not work?
claude_desktop_config.json cannot take an HTTP entry with a headers
block. That file starts local programs, so an entry describing a remote
HTTP server with an Authorization header is ignored, and the app reports
only that some servers failed to load. Nothing names the cause.
The snippet that looks as if it should work is an editor's config, which is a different format for a different mechanism. Claude Desktop reaches a remote server in two ways: a custom connector, above, or a local bridge program, below.
Can I connect Claude with a token instead?
Yes, through a local bridge, and it is rarely needed. It suits a credential the
connector cannot give you, such as a read-only token. The bridge runs
mcp-remote as a local program that relays to RankX AI:
{ "mcpServers": { "rankxai": { "command": "cmd", "args": [ "/c", "npx", "-y", "mcp-remote", "https://app.rankxai.com/api/mcp", "--header", "Authorization:${RANKXAI_TOKEN}" ], "env": { "RANKXAI_TOKEN": "Bearer rxai_your_token_here" } } }}Two Windows details in that block prevent a silent failure:
cmd /cwrapsnpx. On Windows,npxis a shell script rather than an executable, so a config that names it directly cannot start it.- The credential sits in
env, and the header value has no space.mcp-remotesplits--headeron the first space, so"Authorization: Bearer rxai_..."arrives cut short. Referencing an environment variable keeps the argument free of spaces.
On macOS and Linux, use "command": "npx" and drop "cmd" and "/c"; the
header rule still applies. Restart Claude Desktop fully after editing the file.
Some Claude organisations also see a Request headers section in the custom
connector dialog, which Claude offers in beta. There, choose No sign-in and
add an Authorization header whose value is Bearer followed by a space and
your token.
Which capabilities does the connection get?
A Claude connection to RankX AI is granted the full capability set in one decision, not a checkbox per scope. The reasoning is on the authentication page: a partial grant would make a whole group of tools vanish, with the cause a checkbox ticked days earlier.
The safeguards are in how the tools act. Every edit is a dry run first or acts only on something a person approved, every write can be refused, and the site administration capability cannot install a plugin or delete a user.
How do I disconnect Claude?
Either place is enough:
- In Claude, go to Customize, then Connectors, and remove RankX AI. Team and Enterprise Owners remove it in Organization settings, then Connectors.
- In RankX AI, open MCP settings, then Connected apps, and select Disconnect. Use this one if you think the credential has leaked: it takes effect on Claude's next request.
Claude does not let you change a connector's sign-in settings after adding it. To change them, remove the connector and add it again.
Where to go next
- Examples for prompts that work.
- Agent Skills for ready-made workflows.
- Connect Claude Code for the command line.
- Troubleshooting if the connector will not appear.
Questions
Can I add RankX AI on the free Claude plan?
Yes. Claude's Free plan allows one custom connector, and Pro, Max, Team and Enterprise allow more. RankX AI itself includes MCP access on every RankX AI plan and in the trial.
Which OAuth client option should I pick in Claude?
Choose Register automatically. RankX AI supports dynamic client registration, so Claude registers itself and you enter no client ID or secret. If your dialog shows Advanced settings instead, leave them empty.
Does the connector work in Claude Desktop and on claude.ai?
Yes. A custom connector belongs to your Claude account, so once RankX AI is added and signed in it is available in Claude on the web, in the desktop app and on mobile.
How do I turn RankX AI on in a chat?
Select the + button in the chat, choose Connectors and switch RankX AI on for that conversation. Turn it off the same way when you do not need it.
Last updated