MANAGE WITH AI
ZRouter MCP.
Manage your account with AI.
ZRouter includes an MCP server for your account. Connect Claude Code, Cursor, or any MCP client and manage routers, API keys, usage, logs, budgets, and rate limits in plain language.
What is the ZRouter MCP server?
MCP means Model Context Protocol, a way for an AI assistant to use external tools. ZRouter's account MCP server exposes tools for your account, so your assistant can inspect usage and carry out account tasks through an authenticated connection.
Who is it for, and why use it?
Use MCP if you already work in Claude Code, Cursor, or another compatible assistant and want to investigate usage or manage account settings in the same conversation. Use zctl when you want explicit terminal commands or repeatable scripts. Use the dashboard for visual charts and the playground.
Connecting this MCP server does not change your assistant's model provider. It does not enable tool-based model requests on prepaid accounts. See the integration guides for model-request setup and compatibility.
1. Create an MCP token
Open CLI and MCP access in your dashboard, name the token after your client, and select Create MCP token. Copy it when it appears; it is shown once and expires after 30 days. You can revoke it from the same page at any time.
Account API keys only send model requests. They cannot manage your account, so a leaked API key cannot create keys or change budgets.
2. Connect your client
Claude Code:
claude mcp add --transport http zrouter https://zrouter.si/account/mcp \
--header "Authorization: Bearer YOUR_MCP_TOKEN"Cursor: add the server to your user-level ~/.cursor/mcp.json. Replace the token placeholder with your MCP token:
{
"mcpServers": {
"zrouter": {
"url": "https://zrouter.si/account/mcp",
"headers": { "Authorization": "Bearer YOUR_MCP_TOKEN" }
}
}
}Keep tokens out of shared repository configuration. Client formats differ: this Cursor example uses mcpServers; VS Code has its own configuration format. See the official Claude Code, Cursor, and VS Code MCP instructions when configuring or upgrading a client.
3. Ask for what you need
- “How much did I spend this week, by model?”
- “Show the requests that failed with 429 today.”
- “Create a router called support-bot that splits traffic between two models.”
- “Create an API key for staging and cap my spend at $10 a day.”
Available tools
| Area | Tools |
|---|---|
| Account | account_info, credit_ledger |
| Usage | usage_summary, usage_daily, usage_by_model, usage_log |
| Logs | logs_search, logs_get, logs_stats |
| Routers | models_list, virtual_models_list, virtual_models_upsert, virtual_models_delete |
| API keys | api_keys_list, api_keys_create, api_keys_delete |
| Budgets | budgets_list, budgets_upsert, budgets_delete, budgets_reset |
| Rate limits | rate_limits_list, rate_limits_upsert, rate_limits_delete, rate_limits_reset |
Frequently asked questions
Does an account API key work for MCP?
No. Create a separate MCP management token under CLI and MCP access. Model-request API keys cannot manage your account.
Can the assistant manage another account?
No. Each tool runs within the authenticated account. Enabled budgets and rate limits follow the same account rules as the dashboard.
What if the token expires or I stop using the assistant?
Create a new token after the 30-day expiry and update your client configuration. Revoke tokens you no longer need from CLI and MCP access.
Safety
The MCP server can only see and change your own account, with the same rules as your dashboard. Deletes and resets are marked as destructive, so clients that confirm risky actions will ask you first. A new API key's secret appears in the tool result, and so in your agent's history; treat that history accordingly.