Skip to content

Connect Cursor to Elaichi once. From then on, every app your organization has connected in Elaichi — and every tool you are allowed to use — is reachable from Cursor.

Where to find the endpoint: in the Elaichi web app, open Connect your AI client and copy the endpoint shown there.

Elaichi works the same way behind every client: one connection, one consent screen, your own access. How Elaichi connects to AI clients covers that part — this page is the Cursor half.

Before you start

Requirement Detail
Cursor Any recent version with MCP support
Where it applies Cursor reads a file on your machine, so this is per-machine — and optionally per-project. Unlike Claude and ChatGPT, there is nothing account-level to configure
Elaichi account The same sign-in you use for the web app

Connect Cursor to Elaichi

  1. Open ~/.cursor/mcp.json, or .cursor/mcp.json for one project only.
  2. Add an entry under mcpServers with the endpoint as its url.
  3. Reload Cursor, then sign in to Elaichi and approve the access.

The entry is just a name and a URL:

{
  "mcpServers": {
    "elaichi": {
      "url": "https://api.elaichi.ai/mcp"
    }
  }
}

Use the endpoint exactly as Elaichi gives it to you, in place of the URL above. If the file already has other servers under mcpServers, add this alongside them rather than replacing the block.

Info

There is no type field and no command to run. Cursor detects the transport from the endpoint itself, and Elaichi supports dynamic client registration — so you do not need an auth block, a client ID, or a secret.

Reloading brings up Elaichi's sign-in, then the consent screen. Only Read is granted by default, and for most Cursor use you want Read and Run tools — see Approve what the client can do for what each scope covers.

Connect your first app

Start by pasting this into Cursor — it is the same prompt the Connect your AI client dialog hands you, and it both confirms the connection and gets you moving:

Elaichi is connected. List the Elaichi tools you have, then help me connect my first app. If a sign-in comes up, stop and tell me. I approve the access myself.

Those last two sentences are deliberate. A model cannot grant itself access to Elaichi — approving is yours to do. Without them a client will happily report success while the sign-in sits there unanswered.

Connecting the app itself happens in Elaichi, not in Cursor — see Connecting an account.

Anything you connect appears in Cursor without editing mcp.json again — same endpoint, more tools behind it. The same is true in reverse: disconnect in Elaichi and it disappears from Cursor.

Good to know

  • Global or per-project. ~/.cursor/mcp.json applies everywhere; .cursor/mcp.json in a repository applies to that project only. Use the project file when only one codebase should reach your tools.
  • Per-machine. Setting this up on your laptop does not set it up on your desktop — repeat it on each machine, and sign in on each.
  • To disconnect, remove the entry from mcp.json and reload, or revoke the grant in Elaichi — revoking stops the token working immediately.

Troubleshooting

Symptom Likely cause
The server never appears after editing the file Invalid JSON — a trailing comma or a missing brace. Check the file parses, then reload Cursor
No sign-in prompt appears Cursor did not reload. Reload the window, then open the MCP settings to trigger the connection
No tools appear after signing in No apps connected in Elaichi yet, or no toolbox you can reach. Connect an app first
A tool you expect is missing It is blocked by a restriction, or it is not in a toolbox you can reach. Connected tools are never listed, so ask Cursor to search for it
A call is refused after consent The scope it needs was not granted. Reconnect and widen the consent