ContextStream MCP Server

import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'; import { z } from 'zod'; export function registerPrompts(server: McpServer) { // Code exploration prompt server.registerPrompt( 'explore-codebase', { title: 'Explore Codebase', description: 'Get an overview of a project codebase structure and key components', argsSchema: { project_id: z.string().uuid().describe('Project ID to explore'), focus_area: z.string().optional().describe('Optional area to focus on (e.g., "authentication", "api routes")'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `I want to understand the codebase for project ${args.project_id}${args.focus_area ? ` with focus on ${args.focus_area}` : ''}. Please help me by: 1. First, use \`projects.overview\` to get the project summary 2. Use \`projects.files\` to list the key files 3. Use \`search.semantic\` to find relevant code${args.focus_area ? ` related to "${args.focus_area}"` : ''} 4. Summarize the architecture and key patterns you observe Provide a clear, structured overview that helps me navigate this codebase effectively.`, }, }, ], }) ); // Memory capture prompt server.registerPrompt( 'capture-decision', { title: 'Capture Decision', description: 'Document an architectural or technical decision in workspace memory', argsSchema: { workspace_id: z.string().uuid().describe('Workspace ID'), decision_title: z.string().describe('Brief title of the decision'), context: z.string().describe('What prompted this decision'), decision: z.string().describe('The decision made'), consequences: z.string().optional().describe('Expected consequences or tradeoffs'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Please document the following decision in workspace memory: **Title:** ${args.decision_title} **Context:** ${args.context} **Decision:** ${args.decision} ${args.consequences ? `**Consequences:** ${args.consequences}` : ''} Use \`memory.create_event\` with: - event_type: "decision" - workspace_id: "${args.workspace_id}" - title: The decision title - content: A well-formatted ADR (Architecture Decision Record) with context, decision, and consequences - metadata: Include relevant tags and status After creating, confirm the decision was recorded and summarize it.`, }, }, ], }) ); // Code review context prompt server.registerPrompt( 'review-context', { title: 'Code Review Context', description: 'Build context for reviewing code changes', argsSchema: { project_id: z.string().uuid().describe('Project ID'), file_paths: z.string().describe('Comma-separated file paths being changed'), change_description: z.string().describe('Brief description of the changes'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `I need context to review changes in these files: ${args.file_paths} Change description: ${args.change_description} Please help me understand the impact by: 1. Use \`graph.dependencies\` to find what depends on these files 2. Use \`graph.impact\` to analyze potential impact 3. Use \`memory.search\` to find related decisions or notes about these areas 4. Use \`search.semantic\` to find related code patterns Provide: - Summary of what these files do - What other parts of the codebase might be affected - Any relevant past decisions or context from memory - Potential risks or areas to focus the review on`, }, }, ], }) ); // Debug investigation prompt server.registerPrompt( 'investigate-bug', { title: 'Investigate Bug', description: 'Build context for debugging an issue', argsSchema: { project_id: z.string().uuid().describe('Project ID'), error_message: z.string().describe('Error message or symptom'), affected_area: z.string().optional().describe('Known affected area or component'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `I'm investigating a bug: **Error/Symptom:** ${args.error_message} ${args.affected_area ? `**Affected Area:** ${args.affected_area}` : ''} Please help me investigate by: 1. Use \`search.semantic\` to find code related to this error 2. Use \`search.pattern\` to find where similar errors are thrown 3. Use \`graph.call_path\` to trace call flows if we identify key functions 4. Use \`memory.search\` to check if this issue has been encountered before Provide: - Likely locations where this error originates - Call flow analysis - Any related past issues from memory - Suggested debugging approach`, }, }, ], }) ); // Knowledge graph exploration prompt server.registerPrompt( 'explore-knowledge', { title: 'Explore Knowledge Graph', description: 'Navigate and understand the knowledge graph for a workspace', argsSchema: { workspace_id: z.string().uuid().describe('Workspace ID'), starting_topic: z.string().optional().describe('Topic to start exploration from'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Help me explore the knowledge captured in workspace ${args.workspace_id}${args.starting_topic ? ` starting from "${args.starting_topic}"` : ''}. Please: 1. Use \`memory.list_nodes\` to see available knowledge nodes 2. Use \`memory.decisions\` to see decision history 3. ${args.starting_topic ? `Use \`memory.search\` to find nodes related to "${args.starting_topic}"` : 'Use \`memory.summary\` to get an overview'} 4. Use \`graph.related\` to explore connections between nodes Provide: - Overview of knowledge captured - Key themes and topics - Important decisions and their rationale - Connections between different pieces of knowledge`, }, }, ], }) ); // Onboarding context prompt server.registerPrompt( 'onboard-to-project', { title: 'Project Onboarding', description: 'Generate onboarding context for a new team member', argsSchema: { project_id: z.string().uuid().describe('Project ID'), workspace_id: z.string().uuid().describe('Workspace ID'), role: z.string().optional().describe('Role of the person being onboarded (e.g., "backend developer", "frontend developer")'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Create an onboarding guide for a new team member joining this project. ${args.role ? `They will be working as a ${args.role}.` : ''} Please gather comprehensive context: 1. Use \`projects.overview\` and \`projects.statistics\` for project summary 2. Use \`projects.files\` to identify key entry points 3. Use \`memory.timeline\` to see recent activity and changes 4. Use \`memory.decisions\` to understand key architectural choices 5. Use \`search.semantic\` to find documentation and READMEs Provide an onboarding guide that includes: - Project overview and purpose - Technology stack and architecture - Key files and entry points${args.role ? ` relevant to ${args.role}` : ''} - Important decisions and their rationale - Recent changes and current focus areas - Getting started steps`, }, }, ], }) ); // Refactoring analysis prompt server.registerPrompt( 'analyze-refactoring', { title: 'Refactoring Analysis', description: 'Analyze a codebase for refactoring opportunities', argsSchema: { project_id: z.string().uuid().describe('Project ID'), target_area: z.string().optional().describe('Specific area to analyze'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Analyze the codebase for refactoring opportunities${args.target_area ? ` in ${args.target_area}` : ''}. Please investigate: 1. Use \`graph.circular_dependencies\` to find circular dependencies 2. Use \`graph.unused_code\` to find dead code 3. Use \`search.pattern\` to find code duplication patterns 4. Use \`projects.statistics\` to identify complex areas Provide: - Circular dependencies that should be broken - Unused code that can be removed - Duplicate patterns that could be consolidated - High complexity areas that need simplification - Prioritized refactoring recommendations`, }, }, ], }) ); // AI context building prompt server.registerPrompt( 'build-context', { title: 'Build LLM Context', description: 'Build comprehensive context for an LLM task', argsSchema: { query: z.string().describe('What you need context for'), workspace_id: z.string().uuid().optional().describe('Workspace ID'), project_id: z.string().uuid().optional().describe('Project ID'), include_memory: z.string().optional().describe('Include memory/decisions ("true" or "false")'), }, }, async (args) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Build comprehensive context for the following task: **Query:** ${args.query} Please use \`ai.enhanced_context\` with: - query: "${args.query}" ${args.workspace_id ? `- workspace_id: "${args.workspace_id}"` : ''} ${args.project_id ? `- project_id: "${args.project_id}"` : ''} - include_code: true - include_docs: true - include_memory: ${args.include_memory ?? true} Then synthesize the retrieved context into a coherent briefing that will help with the task.`, }, }, ], }) ); }