Connect Murder at the Prompt

Copy this prompt into your coding agent to connect it to the case authoring MCP server.

Endpoint: https://mcp.murder-at-the-prompt.aufdemdev.spaceport.host/mcp

Installation prompt

Install the Murder at the Prompt remote MCP server in this client's MCP configuration.

Server name: murder-at-the-prompt
Endpoint: https://mcp.murder-at-the-prompt.aufdemdev.spaceport.host/mcp
Transport: Streamable HTTP
Authentication: OAuth 2.0 with S256 PKCE and administrator consent.
Requested scopes: cases:read cases:write

Use this client's supported remote MCP configuration. Preserve existing servers. This server requires a pre-registered public OAuth client and does not support dynamic client registration. Determine this client's exact OAuth callback URL and ask the server administrator to register it and provide the client ID. Do not invent a client ID or callback URL. If this client cannot use a pre-registered client ID, explain that limitation before changing configuration. Complete sign-in through the browser; never request passwords or tokens in chat.

After connecting, list the tools and verify access with list_cases. For new cases use begin_case_creation with mode="agent", then get_creation_stage and submit_creation_stage so the calling agent does the writing. Do not request cases:ai or invoke backend OpenAI tools unless I explicitly authorize that spending. The calling agent's own model usage still applies.

An administrator must register your client's exact OAuth callback URL before sign-in. The prompt requests read/write access and agent-powered creation. Backend OpenAI spending requires separate authorization.