Clear Thought 1.5

import fs from "node:fs"; import path from "node:path"; import { Server } from "@modelcontextprotocol/sdk/server/index.js"; import { CallToolRequestSchema, ErrorCode, ListResourcesRequestSchema, ListResourceTemplatesRequestSchema, ListToolsRequestSchema, McpError, ReadResourceRequestSchema, } from "@modelcontextprotocol/sdk/types.js"; import { zodToJsonSchema } from "zod-to-json-schema"; import { SessionState } from "./state/SessionState.js"; import { ClearThoughtParamsSchema, handleClearThoughtTool, } from "./tools/index.js"; import { parseSrcbook, srcbookToResource, } from "./utils/srcbookParser.js"; /** * Creates a Clear Thought MCP server instance for a specific session * @param sessionId - Unique identifier for this session * @param config - Server configuration * @returns Server instance configured for this session */ export default function createClearThoughtServer({ sessionId, config, }) { // Create a new low-level MCP Server instance for each session const server = new Server({ name: "clear-thought", version: "0.0.5", }, { capabilities: { tools: { listChanged: true }, resources: { listChanged: true }, }, }); // Initialize session state const sessionState = new SessionState(sessionId, config); // Register request handlers for tools/list and tools/call server.setRequestHandler(ListToolsRequestSchema, () => ({ tools: [ { name: "clear_thought", title: "Clear Thought", description: "Unified Clear Thought reasoning tool. Operations: sequential_thinking (chain or patterns via pattern/patternParams), mental_model, debugging_approach, creative_thinking, visual_reasoning, metacognitive_monitoring, scientific_method, collaborative_reasoning, decision_framework, socratic_method, structured_argumentation, systems_thinking, research, analogical_reasoning, causal_analysis, statistical_reasoning, simulation, optimization, ethical_analysis, visual_dashboard, custom_framework, code_execution, tree_of_thought, beam_search, mcts, graph_of_thought, orchestration_suggest. See resource guide://clear-thought-operations for details.", inputSchema: zodToJsonSchema(ClearThoughtParamsSchema, { strictUnions: true, }), }, ], })); server.setRequestHandler(CallToolRequestSchema, async (request) => { const { name, arguments: toolArgs } = request.params; // Handle clear_thought tool if (name === "clear_thought") { const parse = await ClearThoughtParamsSchema.safeParseAsync(toolArgs ?? {}); if (!parse.success) { throw new McpError(ErrorCode.InvalidParams, `Invalid arguments for clear_thought: ${parse.error.message}`); } const result = await handleClearThoughtTool(sessionState, parse.data); return result; } throw new McpError(ErrorCode.InvalidParams, `Tool ${name} not found`); }); // Discover available notebooks const notebooksDir = path.resolve(process.cwd(), "../srcbook-examples"); const notebooks = []; const notebookContents = new Map(); try { if (fs.existsSync(notebooksDir)) { const files = fs.readdirSync(notebooksDir); for (const file of files) { if (file.endsWith(".src.md")) { const filePath = path.join(notebooksDir, file); const contents = fs.readFileSync(filePath, "utf-8"); const notebookName = file.replace(".src.md", ""); try { const parsed = parseSrcbook(contents, file); const resource = srcbookToResource(parsed, notebookName); notebooks.push(resource); notebookContents.set(notebookName, contents); } catch (e) { console.error(`Failed to parse notebook ${file}:`, e); } } } } } catch (e) { console.error("Failed to discover notebooks:", e); } // Expose operation documentation as a resource for transparency const operationsGuideUri = "guide://clear-thought-operations"; // Prefer external markdown file if present; fall back to embedded guide const docsPath = path.resolve(process.cwd(), "docs/clear-thought-operations.md"); let operationsGuideMarkdown = `# Clear Thought Operations Guide\n\nThis server exposes a single tool 'clear_thought' with many operations.\n\n- sequential_thinking: Chain-of-thought. Optional pattern selection via parameters.pattern ('chain'|'tree'|'beam'|'mcts'|'graph'|'auto') and parameters.patternParams (pattern-specific settings).\n- mental_model: Apply a mental model (parameters: model, steps, reasoning, conclusion).\n- debugging_approach: Structured debugging (parameters: approach, steps, findings, resolution).\n- creative_thinking: Idea generation (parameters: ideas, techniques, connections, insights, iteration, nextIdeaNeeded).\n- visual_reasoning: Diagram operations (parameters: diagramId, diagramType, iteration, nextOperationNeeded).\n- metacognitive_monitoring: Monitor reasoning (parameters: stage, overallConfidence, uncertaintyAreas, recommendedApproach, iteration, nextAssessmentNeeded).\n- scientific_method: Inquiry workflow (parameters: stage, iteration, nextStageNeeded).\n- collaborative_reasoning: Multi-persona thinking (parameters: personas, contributions, stage, activePersonaId, iteration, nextContributionNeeded).\n- decision_framework: Options, criteria, outcomes (parameters: options, criteria, stakeholders, constraints, analysisType, stage, iteration, nextStageNeeded).\n- socratic_method: Question-driven argumentation (parameters: claim, premises, conclusion, argumentType, confidence, stage, iteration, nextArgumentNeeded).\n- structured_argumentation: Formal arguments (parameters: premises, conclusion, argumentType, confidence, respondsTo, supports, contradicts, strengths, weaknesses, relevance, iteration, nextArgumentNeeded).\n- systems_thinking: System mapping (parameters: components, relationships, feedbackLoops, emergentProperties, leveragePoints, iteration, nextAnalysisNeeded).\n- research: Returns structured placeholders for findings/citations (parameters: none defined).\n- analogical_reasoning: Map between domains (parameters: sourceDomain, targetDomain, mappings, inferredInsights).\n- causal_analysis: Causal graphs and interventions (parameters: graph, intervention, predictedEffects, counterfactual, notes).\n- statistical_reasoning: modes: summary|bayes|hypothesis_test|monte_carlo (parameters vary by mode).\n- simulation: Simple simulation shell (parameters: steps).\n- optimization: Simple optimization shell (parameters: none defined).\n- ethical_analysis: Evaluate with a framework (parameters: framework, score?).\n- visual_dashboard: Dashboard skeleton (parameters: panels, layout, refreshRate).\n- custom_framework: Define custom stages/rules/metrics (parameters: stages, rules, metrics).\n- code_execution: Restricted; only Python when enabled (parameters: language, code).\n- tree_of_thought | beam_search | mcts | graph_of_thought: Pattern-specific structures for exploration (parameters are pattern-specific).\n- orchestration_suggest: Suggests tool combinations (parameters: none defined).\n\nNote: Many operations accept a generic 'parameters' object; see field names above. For 'sequential_thinking', you can choose patterns or set pattern: 'auto' to let the server select based on prompt/params.`; try { if (fs.existsSync(docsPath)) { operationsGuideMarkdown = fs.readFileSync(docsPath, "utf-8"); } } catch { } server.setRequestHandler(ListResourcesRequestSchema, async () => ({ resources: [ { uri: operationsGuideUri, name: "Clear Thought Operations", description: "Documentation for all clear_thought operations and parameters", mimeType: "text/markdown", }, ...notebooks, // Add a guide for notebook interaction { uri: "guide://notebook-interaction", name: "Notebook Interaction Guide", description: "Instructions for working with Srcbook notebooks in MCP", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.9, }, }, // Example data resources for each operation { uri: "examples://systems-thinking", name: "Systems Thinking Examples", description: "Example parameter structures for systems thinking analysis", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://decision-framework", name: "Decision Framework Examples", description: "Multi-criteria and expected utility decision examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://causal-analysis", name: "Causal Analysis Examples", description: "Causal graph structures with intervention examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://creative-thinking", name: "Creative Thinking Examples", description: "SCAMPER and brainstorming technique examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://scientific-method", name: "Scientific Method Examples", description: "Hypothesis testing and experiment structure examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://simulation", name: "Simulation Examples", description: "Discrete-time simulation model examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://optimization", name: "Optimization Examples", description: "Resource allocation and optimization examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://ethical-analysis", name: "Ethical Analysis Examples", description: "Ethical framework analysis examples", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.8, }, }, { uri: "examples://visual-dashboard", name: "Visual Dashboard Examples", description: "Interactive dashboard creation with MCP UI patterns", mimeType: "text/markdown", annotations: { audience: ["assistant"], priority: 0.9, }, }, ], })); // Add resource templates for dynamic notebook access server.setRequestHandler(ListResourceTemplatesRequestSchema, async () => ({ resourceTemplates: [ { uriTemplate: "notebook:///{name}", name: "Srcbook Notebooks", title: "📓 Interactive Notebooks", description: "Access Srcbook notebooks by name for interactive TypeScript/JavaScript execution", mimeType: "text/markdown", }, { uriTemplate: "notebook:///{name}#cell-{index}", name: "Notebook Cells", title: "📝 Specific Notebook Cell", description: "Access a specific cell within a Srcbook notebook", mimeType: "text/plain", }, ], })); server.setRequestHandler(ReadResourceRequestSchema, async (request) => { const { uri } = request.params; // Handle operations guide if (uri === operationsGuideUri) { return { contents: [{ uri: operationsGuideUri, mimeType: "text/markdown", text: operationsGuideMarkdown }], }; } // Handle example resources if (uri.startsWith("examples://")) { const exampleType = uri.replace("examples://", ""); // Try multiple possible locations for the resources const possiblePaths = [ path.join(process.cwd(), `dist/resources/examples/${exampleType}.md`), // dist from cwd path.join(process.cwd(), `src/resources/examples/${exampleType}.md`), // src from cwd path.join(process.cwd(), `resources/examples/${exampleType}.md`), // resources from cwd ]; for (const examplePath of possiblePaths) { try { if (fs.existsSync(examplePath)) { const content = fs.readFileSync(examplePath, "utf-8"); return { contents: [{ uri: uri, mimeType: "text/markdown", text: content }], }; } } catch (error) { // Continue to next path } } throw new McpError(ErrorCode.InvalidParams, `Example resource not found: ${uri}. Tried paths: ${possiblePaths.join(", ")}`); } // Handle notebook interaction guide if (uri === "guide://notebook-interaction") { const interactionGuide = `# Notebook Interaction Guide ## Working with Srcbook Notebooks in MCP Srcbook notebooks (.src.md files) are interactive documents that combine: - **Markdown cells** for documentation and explanations - **Code cells** with JavaScript/TypeScript that can be executed - **Package.json** for dependencies - **tsconfig.json** for TypeScript configuration (when applicable) ## How to Execute Code Cells When you encounter a code cell in a notebook resource: 1. **Identify code cells** - Look for: - Sections starting with \`###### filename.js\` or \`###### filename.ts\` - Standalone code blocks marked as javascript/typescript 2. **Execute the code** using the appropriate tool: - Use \`mcp__ide__executeCode\` for TypeScript/JavaScript execution - Pass the extracted source code from the cell 3. **Process cells sequentially** for educational notebooks to maintain the learning flow ## Example Interaction Pattern \`\`\`typescript // When you see a code cell like this: ###### example.ts console.log("Hello from notebook!"); // Extract and execute: mcp__ide__executeCode({ code: 'console.log("Hello from notebook!");' }) \`\`\` ## Available Notebooks The following notebooks are available as resources: ${notebooks.map((n) => `- **${n.name}** (${n.uri}): ${n.description}`).join("\n")} ## Tips for AI Assistants - Read notebooks sequentially for the best learning experience - Execute code cells to demonstrate concepts interactively - Use the annotations in each resource to understand capabilities - Reference specific cells using URI fragments like \`notebook:///name#cell-3\` `; return { contents: [{ uri: uri, mimeType: "text/markdown", text: interactionGuide }], }; } // Handle notebook resources if (uri.startsWith("notebook:///")) { const notebookName = uri.replace("notebook:///", "").split("#")[0]; const content = notebookContents.get(notebookName); if (!content) { throw new McpError(ErrorCode.InvalidParams, `Notebook not found: ${uri}`); } return { contents: [{ uri: uri, mimeType: "text/markdown", text: content }], }; } throw new McpError(ErrorCode.InvalidParams, `Resource not found: ${uri}`); }); // Return the low-level Server instance return server; } // Re-export config schema so Smithery CLI can adapt a single entry for dev/build/run export { ServerConfigSchema as configSchema } from "./config.js";