supernerve-mcp

# supernerve-mcp Supernerve MCP protocol server and connector scaffold written in TypeScript. The server is intended to expose Supernerve capabilities through Model Context Protocol-compatible tools and handlers. ## Purpose `supernerve-mcp` is the bridge between MCP clients and Supernerve resources. It currently provides a minimal TypeScript server skeleton with tool groups for brain, agents, workflows, and memory. The structure is ready for integration with the final MCP transport, tool schemas, request validation, Supernerve API calls, and resource handling. ## Tech stack - Node.js runtime. - TypeScript with ES modules. - npm scripts for build, clean, start, and typecheck. - MCP-style result objects with text content blocks. ## Repository layout ```text supernerve-mcp/ |-- src/ | |-- server.ts Server entrypoint and startup behavior | |-- handlers/ Shared handler helpers | | `-- index.ts | |-- tools/ Tool groups exposed to MCP clients | | |-- agents.ts | | |-- brain.ts | | |-- memory.ts | | `-- workflows.ts | `-- types.ts Shared MCP result and content types |-- dist/ Compiled output generated by `npm run build` |-- package.json Scripts and package metadata |-- package-lock.json npm lockfile `-- tsconfig.json TypeScript compiler settings ``` ## Tool groups The current scaffold includes placeholder tool groups: - Brain tools for search, retrieval, and knowledge operations. - Agent tools for listing, inspecting, and controlling agents. - Workflow tools for listing, running, and observing workflows. - Memory tools for persistent user or workspace memory operations. Each tool module currently returns a text result through `createTextResult`. Replace placeholders with typed schemas and real request handling as the MCP contract is finalized. ## Prerequisites - Node.js 20 or newer is recommended. - npm is the current package manager. - An MCP client or host will be needed once transport support is implemented. ## Setup ```bash npm install npm run typecheck npm run build npm start ``` ## Available scripts - `npm run build` compiles TypeScript into `dist/`. - `npm run clean` removes compiled output. - `npm start` runs `dist/server.js`. - `npm run typecheck` runs TypeScript without emitting files. ## Local development 1. Install dependencies with `npm install`. 2. Keep shared response helpers in `src/handlers`. 3. Add tool behavior in feature-specific files under `src/tools`. 4. Keep shared MCP types in `src/types.ts`. 5. Run `npm run typecheck`. 6. Build with `npm run build`. 7. Start the compiled server with `npm start`. ## Implementation guidelines - Define explicit input and output schemas for every public tool. - Validate client input before calling Supernerve services. - Keep transport-specific logic out of feature tool modules. - Return structured MCP content instead of ad hoc strings when the client needs machine-readable data. - Make errors actionable and avoid leaking secrets. ## Testing and validation This scaffold currently has typechecking but no test script. At minimum run: ```bash npm run typecheck npm run build ``` Add tests around tool schema validation, handler behavior, and API integration as the server grows. ## Contribution notes - Keep TypeScript strict and MCP contracts explicit. - Prefer feature-owned tools with shared helpers for common response behavior. - Keep secrets and credentials out of source control. - Update this README when setup, transport behavior, tool schemas, environment variables, or scripts change.