> ## Documentation Index > Fetch the complete documentation index at: https://docs.slickerhq.com/llms.txt > Use this file to discover all available pages before exploring further. # Slicker for AI assistants > Explore your billing, subscriptions, payments, and alerts with Slicker MCP Connect Slicker to an AI assistant and ask questions about your business in your own words. Slicker MCP combines your data with Slicker's billing and payments expertise, so your assistant can investigate a change, explain what the evidence shows, and suggest relevant best practices. It can also help you monitor for problems and investigate alerts. Slicker hosts the MCP server. You do not need to install or run a Slicker server yourself. ## Get started with your agent Use an AI assistant that supports remote MCP connections. Copy this instruction into the assistant you want to use—Claude, Claude Code, Codex, ChatGPT, or another compatible client. It should identify the client, guide you through Slicker sign-in, and perform a read-only check. You do not need to create or paste an API key to use the recommended OAuth connection. Help me connect Slicker's hosted MCP server to the AI client I am using. First read [https://docs.slickerhq.com/mcp/connect](https://docs.slickerhq.com/mcp/connect) and follow its current compatibility and setup guidance for this specific client. If you cannot open that page, ask me to open it and provide the relevant instructions. Do not assume instructions for one AI client apply to another. The server URL is [https://api.slickerhq.com/mcp](https://api.slickerhq.com/mcp) (remote Streamable HTTP). Use this client's native MCP OAuth sign-in to connect; I do not need to run a local server. First identify this client and its current remote MCP setup flow. If OAuth sign-in is not available here, explain that before suggesting the documented API-key alternative. Never choose an unauthenticated connection. Prefer the client's native MCP settings or setup command. Use a personal/user-level connection where available, not a repository configuration. If my client requires a workspace administrator to add the connector, explain that step; I should still sign in with my own Slicker account. Make the smallest necessary change and allow long-running analytics calls up to 15 minutes if the client supports a tool timeout. If you cannot configure the client yourself, give me precise steps for it. Guide me through browser sign-in and have me review the permissions I authorize. Use read access for setup when this client allows choosing scopes; alert changes additionally need the alert-write permission. Do not ask me to paste a password, API key, authorization code, or access token into this chat. If OAuth is unavailable and I choose the API-key alternative, explain that the key grants full administrator access and follow the private-key instructions on the connection page. Then verify the connection by listing the Slicker tools and calling list\_recovery\_actions with page\_size=1. An empty successful result is fine. Do not create or change alerts during setup. Tell me what was verified, what I need to do to make the credential available in future sessions, and any access error you encountered. Do not assume optional analytics or alert access merely because the tools are listed. If you prefer to configure the connection yourself, see [manual setup and troubleshooting](/mcp/connect). OAuth sign-in works with compatible remote MCP clients. The [connection guide](/mcp/connect) includes specific steps for Codex, Claude Code, Claude, and ChatGPT, plus an API-key alternative for clients that support private Bearer tokens. ## Examples of what you can do * Get a health check across subscriptions, renewals, authorization, churn, retention, recovery, and revenue at risk. * Investigate a change, such as a drop in payment acceptance or a rise in first-renewal failures, and see which customers or segments are affected. * Explore payment performance by gateway, card network, issuer, country, payment method, 3D Secure outcome, or decline reason. * Examine failed invoices, recovery outcomes, and the exact retries or dunning actions Slicker took. * Review active alerts, investigate why one fired, and see available alert presets. * Describe a billing or payment problem you want to monitor, then review a proposed alert and its backtest before deciding whether to enable it. These are examples, not a fixed menu of questions. Slicker can combine your data with billing and payments expertise, distinguish measured findings from recommendations, and explain data gaps. The available analyses and alert signals depend on your connected data and enabled Slicker capabilities. ## What happens when you connect The connection uses your Slicker account and the permissions you authorize. OAuth read access does not authorize alert changes: those additionally require the alert-write permission. Asking a question or investigating an alert is read-only. Alert proposals and rule changes are different: a proposal is saved but not enabled, while enabling a rule can start evaluations and notifications. If you use the API-key alternative, treat that key as a full administrator credential. Before your assistant creates or changes an alert, review the exact change and use your MCP client's tool-approval controls. The assistant's promise to ask first is helpful, but it is not an access control. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.