Give Cursor an email inbox

Cursor's agent lives in your editor, which is exactly where you are when a build breaks or a customer replies about the bug you are fixing. Anjal gives it an inbox so it can read that mail and answer it without you switching windows.

Setup is one JSON entry.

Connect it

Create or edit .cursor/mcp.json in the project, or ~/.cursor/mcp.json to have it everywhere:

{
  "mcpServers": {
    "anjal": {
      "url": "https://mcp.anjal.app/mcp"
    }
  }
}

That is the whole file. There is no API key in it, which is the point: Anjal's MCP server implements the OAuth flow, so Cursor registers itself and sends you to Anjal to sign in and click Allow. Nothing secret lands in a file you might commit.

Open the MCP section of Cursor's settings and you should see anjal with its tools listed. If it is not connected yet, Cursor will prompt for sign-in the first time the agent uses a tool.

Project file or global file? Put it in .cursor/mcp.json and commit it when the whole team should get the same server; each person still signs in as themselves, so they see their own inboxes. Put it in ~/.cursor/mcp.json when it is yours alone.

What Cursor gets

47 tools. The ones you will actually use from an editor:

What it is good for in an editor

Close the loop on a bug report. Ask the agent to find the thread where the customer described the crash, read it alongside the file it points at, and draft the reply once the fix is in. It has both halves in one context, which is the reason to do this in the editor rather than in a browser tab.

Sign your test builds up for things. An agent wiring up a third-party API usually needs an account on it. With an inbox it can register, receive the confirmation mail and pull the verification code itself.

Keep a per-project address. create_inbox gives each client project its own address on your domain, so the mail about that project lives with that project.

The part worth reading before you connect it

Inbound mail is text written by strangers, heading into the context of an agent that edits your code. Treat it as hostile, because it is.

Anjal screens every inbound message for prompt injection on the ingest path and quarantines what it judges hostile. Quarantined mail is not returned by list_messages or get_thread, so the agent never sees it. A person releases it from the Anjal web app; an agent cannot release its own quarantine, because the obvious attack is to ask it to. We publish the architecture and not the rules. Read the honest limits too: detection is advisory, not a trust boundary.

On by default, nothing to configure.

Limiting what the agent can reach

The OAuth connector covers your whole account. To scope Cursor to a single workspace, run the local server with a workspace key instead:

{
  "mcpServers": {
    "anjal": {
      "command": "npx",
      "args": ["-y", "anjal-mcp@latest"],
      "env": { "MYAGENTMAIL_KEY": "wk_..." }
    }
  }
}

A workspace key cannot list, read or send from any other workspace. If you use this form, keep it in ~/.cursor/mcp.json rather than the committed project file, because now there is a real credential in it.

Cost

Starter is $5/month for 2 inboxes and 6,000 emails. Developer is $15/month for 10 inboxes and 20,000 emails. Extra inboxes are $2/month each. 7-day trial.

Next

• Connect Anjal to Claude Code
• Connect Anjal to Codex
• MCP server reference and agent safety
• WebSocket vs webhooks for agent email

Config format verified against Cursor's MCP documentation on 2026-09-28.