Report Abuse

Basic Information

Cipher is an open source memory layer designed specifically for coding agents and developer IDE workflows. It provides a persistent, shared memory service that integrates with MCP-compatible clients and IDEs so coding assistants retain context across sessions and tools. The project targets integration with Cursor, Windsurf, Claude Desktop, Claude Code, Gemini CLI, AWS Kiro, VS Code, Roo Code and other MCP clients. Cipher can run as a CLI tool, an MCP server, an API server or a web UI, and is distributed via npm, Docker and source builds. Configuration is driven by a main YAML file and environment variables, and the project documents LLM provider configuration, embedding and vector store options, chat history backends, workspace memory and MCP integration. The README emphasizes use by developer teams and coding agents to capture and share code-related knowledge and reasoning.

Links

Categorization

App Details

Features
Cipher focuses on memory and integration features for coding agents. It offers MCP integration to connect with many IDEs and clients. It implements a dual memory layer distinguishing programming concepts and past interactions from model reasoning traces. Built-in tools support semantic memory operations, reasoning extraction and evaluation, workspace/team memory capture, knowledge graph node and edge operations, and a bash execution system tool. Vector store support includes qdrant, milvus and in-memory options while chat history can use SQLite or PostgreSQL. Multiple LLM providers are supported such as OpenAI, Anthropic, Qwen and others with configurable embedding providers. Deployment and usage options include npm packages, Docker compose, running from source, CLI modes (interactive, API, MCP, UI), and configuration via memAgent/cipher.yml and .env variables.
Use Cases
Cipher helps developer teams and coding agent creators maintain persistent, context-rich memory that scales with a codebase and spans IDEs and tools. By auto-generating and storing coding memories and structured reasoning traces, it preserves past interactions and model thought steps to improve continuity across sessions. Teams can share workspace memory in real time and search project knowledge for faster onboarding, debugging and feature work. MCP server mode allows existing MCP-enabled coding assistants to use Cipher as a centralized memory backend. Integration with vector stores and chat history backends enables semantic search and retrieval. The CLI, Docker and npm distribution make it straightforward to run locally or in CI, and documentation, examples and built-in tools help implement memory-driven coding workflows.

Please fill the required fields*