Cipher serves as a specialized MCP server designed to provide persistent memory management for coding agents and AI assistants. When cipher runs as an MCP server, it can operate in two distinct modes depending on your configuration, each offering different capabilities and tool access patterns. Cipher MCP Server Architecture - Browse and connect to various MCP servers including File System, GitHub, Git, Memory, and SQLite

​
MCP Server Modes

Cipher offers two operational modes when running as an MCP server, each designed for different use cases and integration patterns.

​
Default Mode

​
General Idea & Why Use It

The default mode provides a streamlined, unified interface focused purely on cipher’s core memory management capabilities. This mode is ideal when you want:
  • Clean Integration: Simple, single-tool interface that doesn’t clutter your development environment
  • Memory-Focused Workflow: Direct access to cipher’s persistent memory without additional complexity
  • Lightweight Setup: Minimal configuration with maximum memory benefits
  • Consistent Experience: Unified tool interface that handles all memory operations seamlessly

​
Diagram

In default mode, cipher exposes only the ask_cipher tool, which serves as a unified interface for all memory operations. External coding agents connect directly to cipher’s memory capabilities without access to additional MCP tools. Cipher Storing Memory Cipher Retrieving Memory

​
The Ask Cipher Tool

When cipher runs in default mode, it exposes a single, powerful tool that coding agents can use for comprehensive memory management:

​
Storage Operations

  • Knowledge Storage: Store code patterns, solutions, and best practices
  • Reflection Memory: Record learning experiences and improvements
  • Entity Management: Add coding concepts and their relationships
  • Context Preservation: Maintain project and codebase understanding

​
Retrieval Operations

  • Semantic Search: Find relevant past solutions and patterns
  • Context Retrieval: Get background information for current coding tasks
  • Pattern Recognition: Identify recurring themes and successful approaches
  • Cross-Session Continuity: Access knowledge from previous sessions

​
Aggregator Mode

​
General Idea & Why Use It

The aggregator mode transforms cipher into a comprehensive development hub that combines its memory capabilities with external MCP tools. This mode is perfect when you need:
  • Comprehensive Toolset: Access to both cipher’s memory tools and external MCP servers (filesystem, databases, web browsing, etc.)
  • Unified Agent Experience: Single connection point that provides access to multiple development tools
  • Advanced Workflows: Complex development tasks that require both memory management and external tool integration
  • Centralized Development Hub: One MCP server that aggregates multiple capabilities

​
Diagram

In aggregator mode, cipher acts as a proxy that exposes both its internal memory tools and tools from configured MCP servers. External coding agents connect to cipher and gain access to the full ecosystem of development tools. Cipher Aggregator Mode in Cursor

​
Available Tools

​
Cipher’s Internal Tools

  • ask_cipher: Unified memory interface (same as default mode)
  • cipher_memory_search: Direct semantic search capabilities
  • cipher_store_reasoning_memory: Advanced reasoning pattern storage
  • cipher_extract_entities: Knowledge graph entity extraction
  • cipher_query_graph: Advanced graph-based queries
  • Plus all other internal memory and reasoning tools

​
External MCP Tools

When configured with mcpServers in your cipher.yml, you also get access to:
  • Filesystem Tools: File operations, directory navigation, code analysis
  • Database Tools: SQL queries, schema exploration, data management
  • Web Browsing Tools: Research, documentation lookup, API exploration
  • Git Tools: Repository management, version control operations
  • And More: Any MCP server you configure in your cipher.yml file

​
Mode Selection

The mode cipher operates in depends on your configuration:
  • Default Mode: When cipher is started without additional MCP server configurations
  • Aggregator Mode: When cipher is configured with external MCP servers to proxy through
Both modes provide cipher’s core memory management capabilities, with aggregator mode extending functionality by including external development tools.

​
Memory System Components

Cipher’s memory system consists of three main components that work together to provide comprehensive knowledge storage and retrieval:
  1. Knowledge & Reflection Memory (Vector Store) - Semantic search capabilities for code patterns and solutions
  2. Knowledge Graph (Graph Store) - Relationship modeling between coding concepts and entities
  3. Session & Metadata (SQLite and Postgres Database) - Operational data and conversation history
For detailed information about each memory component, including architecture, configuration options, and use cases, please refer to the Memory Overview page.

​
Next Steps

For detailed setup instructions and configuration guidance, please refer to the Connections page, which provides comprehensive instructions for integrating cipher with your development environment. The connections page covers everything you need to get cipher running as your persistent memory layer, transforming your coding agent into a persistent coding companion that grows smarter with each interaction.