Claude Server MCP

# Context Management in Claude Server ## Overview The Claude Server implements a sophisticated context management system that handles both project-specific and conversation contexts. This system is designed to maintain context across sessions while organizing information in a structured and retrievable way. ## Directory Structure ``` ~/.claude/ ├── contexts/ # General conversation contexts ├── projects/ # Project-specific contexts │ ├── project-1/ # Contexts for project-1 │ └── project-2/ # Contexts for project-2 └── context-index.json # Quick lookup index for all contexts ``` ## Context Types ### Base Context All contexts share these common properties: ```typescript interface BaseContext { id: string; // Unique identifier content: string; // The actual context content timestamp: string; // ISO timestamp tags?: string[]; // Optional categorization tags metadata?: Record<string, unknown>; // Additional flexible metadata } ``` ### Project Context For project-specific information with relationships: ```typescript interface ProjectContext extends BaseContext { projectId: string; // Project identifier parentContextId?: string; // Optional parent context references?: string[]; // Related context IDs } ``` ### Conversation Context For maintaining conversation continuity: ```typescript interface ConversationContext extends BaseContext { sessionId: string; // Conversation session ID continuationOf?: string; // Previous context ID } ``` ## Tools ### 1. save_project_context Save project-specific context with relationships: ```typescript use_mcp_tool({ server_name: "claude-server", tool_name: "save_project_context", arguments: { id: "feature-design-v1", projectId: "my-project", content: "Feature design discussion...", parentContextId: "requirements-v1", references: ["api-spec-v1"], tags: ["design", "feature"], metadata: { status: "in-progress", priority: "high" } } }); ``` ### 2. save_conversation_context Save conversation context with continuation support: ```typescript use_mcp_tool({ server_name: "claude-server", tool_name: "save_conversation_context", arguments: { id: "chat-2024-01-01-2", sessionId: "session-123", content: "Continuation of previous discussion...", continuationOf: "chat-2024-01-01-1", tags: ["meeting", "planning"], metadata: { participants: ["user1", "user2"] } } }); ``` ### 3. get_context Retrieve context by ID: ```typescript use_mcp_tool({ server_name: "claude-server", tool_name: "get_context", arguments: { id: "feature-design-v1", projectId: "my-project" // Optional for project contexts } }); ``` ### 4. list_contexts List contexts with filtering: ```typescript use_mcp_tool({ server_name: "claude-server", tool_name: "list_contexts", arguments: { projectId: "my-project", // Optional: filter by project tag: "design", // Optional: filter by tag type: "project" // Optional: 'project' or 'conversation' } }); ``` ## Best Practices ### Context Organization 1. Use consistent naming patterns: - Project contexts: `{feature}-{type}-v{version}` - Conversation contexts: `chat-{date}-{sequence}` 2. Tag effectively: - Use project-specific tags - Include status tags - Add type tags (e.g., "design", "meeting") 3. Utilize metadata: - Add status information - Include relevant timestamps - Store participant information - Track versions ### Context Relationships 1. Parent-Child: - Use parentContextId for hierarchical relationships - Link iterations with previous versions 2. References: - Link related contexts - Cross-reference documentation - Connect discussions to implementations ### Session Management 1. Track conversation flows: - Use consistent sessionId for related conversations - Link sequential conversations with continuationOf - Add metadata for conversation status 2. Project Integration: - Tag conversations with project identifiers - Reference project contexts in conversations - Maintain project-specific conversation threads ## Implementation Details ### Storage - Contexts are stored as JSON files - Project contexts are organized in project-specific directories - An index file maintains quick lookup capabilities - All timestamps use ISO 8601 format ### Performance - Lazy loading of contexts - Index-based quick lookups - Efficient filtering mechanisms - Asynchronous file operations ### Security - Contexts are stored in user's home directory - File permissions follow system defaults - No sensitive data in context identifiers - Metadata validation for safe storage ## Example Workflows ### Project Documentation ```typescript // Save initial requirements use_mcp_tool({ server_name: "claude-server", tool_name: "save_project_context", arguments: { id: "requirements-v1", projectId: "my-project", content: "Project requirements...", tags: ["requirements", "initial"] } }); // Save design based on requirements use_mcp_tool({ server_name: "claude-server", tool_name: "save_project_context", arguments: { id: "design-v1", projectId: "my-project", content: "Design specifications...", parentContextId: "requirements-v1", tags: ["design"] } }); ``` ### Conversation Continuity ```typescript // Initial conversation use_mcp_tool({ server_name: "claude-server", tool_name: "save_conversation_context", arguments: { id: "chat-2024-01-01-1", sessionId: "session-123", content: "Initial discussion...", tags: ["meeting"] } }); // Continuation use_mcp_tool({ server_name: "claude-server", tool_name: "save_conversation_context", arguments: { id: "chat-2024-01-01-2", sessionId: "session-123", content: "Continued discussion...", continuationOf: "chat-2024-01-01-1", tags: ["meeting"] } });