Interactive MCP

### 1. Basic Configuration This is a minimal, usable configuration for the **Interactive MCP** server. Assumes the server is started by a simple command like `interactive-mcp` on your PATH. ```json { "mcpServers": { "interactive-mcp": { "command": "interactive-mcp", "args": [], "env": {}, "metadata": { "category": "productivity", "description": "Interactive MCP server for conversational, step-by-step tasks." } } } } ``` **Explanation (not part of JSON):** - `mcpServers`: Top-level key in `~/.claude/settings.json` where all MCP server configs live. - `"interactive-mcp"`: The identifier Claude will use for this server. You’ll reference this name in Claude Code UI. - `command`: The executable that starts the Interactive MCP server. This must be installed and available on your PATH. - `args`: Empty array means no extra arguments; just run `interactive-mcp`. - `env`: Empty object means no special environment variables are needed. - `metadata.category`: Set to `"productivity"` as requested. - `metadata.description`: Optional but helpful for identifying the server in UI. --- ### 2. Advanced Configuration This configuration adds: - A specific Node.js entrypoint - Logging and debug options via env vars - A custom working directory - Timeout and concurrency controls (passed as arguments) Assumes you’re running a Node-based MCP server via `npx` or a local script. ```json { "mcpServers": { "interactive-mcp": { "command": "node", "args": [ "/usr/local/lib/interactive-mcp/server.js", "--max-concurrent-sessions", "5", "--request-timeout-ms", "30000" ], "env": { "INTERACTIVE_MCP_MODE": "development", "INTERACTIVE_MCP_LOG_LEVEL": "debug", "INTERACTIVE_MCP_STORAGE_PATH": "/var/tmp/interactive-mcp", "NODE_ENV": "development" }, "workingDirectory": "/usr/local/lib/interactive-mcp", "metadata": { "category": "productivity", "description": "Interactive MCP server with debug logging and concurrency limits." }, "timeoutSeconds": 60, "disabled": false } } } ``` **Explanation (not part of JSON):** - `command: "node"`: Uses Node.js directly to run the server script. - `args`: - First arg is the path to the MCP server script. - `--max-concurrent-sessions 5`: Limits how many parallel interactive sessions Claude can run. - `--request-timeout-ms 30000`: Server-level timeout per request (30 seconds). - `env`: - `INTERACTIVE_MCP_MODE`: Custom mode flag your server can read. - `INTERACTIVE_MCP_LOG_LEVEL`: Enables verbose logging for debugging. - `INTERACTIVE_MCP_STORAGE_PATH`: Directory for any local state/logs/cache. - `NODE_ENV`: Standard Node environment flag. - `workingDirectory`: Where the process will start; useful if the script uses relative paths. - `metadata.category`: Still `"productivity"`. - `timeoutSeconds`: How long Claude will wait before considering the MCP call failed (client-side). - `disabled`: Explicitly `false` to ensure the server is active. --- ### 3. Use Case Specific Configuration **Use case: “Team Task Orchestration in Production”** This configuration is tailored for a production environment where **Interactive MCP** is used to coordinate multi-step workflows for a team (e.g., managing tasks, checklists, approvals). It emphasizes stability, logging, and a dedicated API endpoint. Assumes: - You deploy the MCP server as a standalone binary `/opt/interactive-mcp/interactive-mcp`. - It connects to a remote backend API for task data. - You want stricter timeouts and production logging. ```json { "mcpServers": { "interactive-mcp": { "command": "/opt/interactive-mcp/interactive-mcp", "args": [ "--mode", "orchestration", "--backend-url", "https://tasks.example.com/api", "--max-concurrent-sessions", "20", "--request-timeout-ms", "15000" ], "env": { "INTERACTIVE_MCP_ENV": "production", "INTERACTIVE_MCP_LOG_LEVEL": "info", "INTERACTIVE_MCP_LOG_FILE": "/var/log/interactive-mcp/interactive-mcp.log", "INTERACTIVE_MCP_API_KEY": "REPLACE_WITH_SECURE_API_KEY", "INTERACTIVE_MCP_ORG_ID": "team-ops", "NODE_ENV": "production" }, "workingDirectory": "/opt/interactive-mcp", "metadata": { "category": "productivity", "description": "Production Interactive MCP for team task orchestration and approvals." }, "timeoutSeconds": 20, "disabled": false } } } ``` **Explanation (not part of JSON):** - `command`: Uses a dedicated production binary. - `args`: - `--mode orchestration`: Tells the server to run in “team workflow” mode (e.g., multi-step tasks, approvals). - `--backend-url https://tasks.example.com/api`: Points to your task/issue backend. - `--max-concurrent-sessions 20`: Higher concurrency for a whole team. - `--request-timeout-ms 15000`: Slightly aggressive server timeout for responsiveness. - `env`: - `INTERACTIVE_MCP_ENV`: Explicitly marks this as production. - `INTERACTIVE_MCP_LOG_LEVEL: "info"`: Less noisy than debug, suitable for production. - `INTERACTIVE_MCP_LOG_FILE`: File path for log aggregation/rotation. - `INTERACTIVE_MCP_API_KEY`: Secret token for the backend; in practice, set this securely (e.g., via OS-level env injection). - `INTERACTIVE_MCP_ORG_ID`: Used by the backend to scope data to a specific team/org. - `NODE_ENV: "production"`: Standard Node/JS convention if the binary is Node-based internally. - `timeoutSeconds: 20`: Claude will consider calls failed if no response within 20 seconds, keeping the UX snappy. - `disabled: false`: Ensures this production configuration is active.