Skip to main content

Connect the Tera MCP connector

The Tera MCP connector is a secure, read-only bridge between your AI assistant (Claude and other MCP-compatible clients) and your Tera finance data. Once connected, you can ask questions in plain English — "what's pending my approval?", "top vendors by spend this quarter", "explain how this expense is coded" — answered with live, tenant-scoped data from your workspace.

What you can ask

  • Transactions & records — "Show expenses over ₹50,000 last month", "List unpaid vendor bills", "Find card transactions at Amazon"
  • Approvals — "What's pending my approval?", "Show this bill's approval chain"
  • Master data — "What GL accounts exist?", "List our expense categories"
  • Context & 'why' — "Explain how this expense is coded", "Who approved this bill?"
  • Analytics & reports — "Top 10 vendors by spend this quarter", "Overdue bills", "Cash-outflow forecast"

Before you start

Access is off by default and granted by your Tera administrator. An admin must enable the connector for your workspace and grant you access (see Integrations → Tera MCP → Manage Access). If you connect without access, sign-in works but the connector reports that access hasn't been granted — contact your admin.

Connect your client

The connector URL is:

https://mcp.tera.cloud/mcp

claude.ai (web)
1. Go to Settings → Connectors.
2. Click Add custom connector.
3. Enter a name (e.g. Tera) and the URL above.
4. Click Add, then Connect — a Tera sign-in window opens.
5. After sign-in, the connector shows Connected and its tools are available in chat.

Claude Desktop — Settings → Connectors → Add custom connector, paste the URL, and complete the Tera sign-in.

Claude Code

claude mcp add --transport http tera https://mcp.tera.cloud/mcp

Run /mcp to sign in, then use the tools in your session.

Other MCP clients — any client that supports remote (streamable-HTTP) MCP servers with OAuth can connect with the same URL. Clients that use a JSON config file usually take this shape:

{
  "mcpServers": {
    "tera": {
      "url": "https://mcp.tera.cloud/mcp"
    }
  }
}

No API key required. Tera uses OAuth — you sign in through your browser the first time you connect, so there's no token or Authorization header to set. If your client asks for one, leave it blank.

If the server seems to be ignored, check the field name your client expects — most use url, but some use serverUrl. Using the wrong one makes the server fail silently.

Signing in

The connector uses your existing Tera login — no separate password. A browser window opens the Tera sign-in page; sign in as usual (email + password, or SSO) and approve the connection. Short-lived tokens refresh automatically, so you rarely need to sign in again. You can disconnect anytime from your client's connector settings.

The connector never sees or stores your password — authentication happens on the Tera login page, exactly like signing into the web app.

Security & privacy

  • Read-only. No tools create, edit, or delete data — your assistant can read and analyze, never modify.
  • Tenant isolation. Your identity comes from your signed-in token on every request, so you only ever see your own workspace's data.
  • Scoped access. Admins grant read-only scopes (transactions, analytics, reports); your token is limited to exactly what you've been granted.
  • No raw queries. The connector offers a fixed, curated set of safe tools — the AI can't run arbitrary database queries — and responses are size-capped.

On AI assistants: data your assistant retrieves is processed by that assistant (e.g. Claude) under its provider's terms. Connect only clients you trust, and disconnect when no longer needed.

Troubleshooting

  • "Access has not been granted" — your admin hasn't enabled the connector for you. Ask them to grant access in Integrations → Tera MCP → Manage Access.
  • Tools don't appear / missing new tools — your client caches the tool list. Toggle the connector off and on (or restart the app); if needed, remove and re-add it.
  • Sign-in window hangs — allow pop-ups for your AI client, complete the Tera login, and approve the connection.
  • A tool says it's not authorized — you don't hold the scope it needs (e.g. analytics). Ask your admin to add it.

Still stuck? See How do I contact support?

Related articles