Cognio

Cognio is a MCP server that provides persistent semantic memory for AI assistants. It stores memories in a SQLite database and offers semantic search across conversations, memories, and projects. With features like LEANN vector search, multilingual capabilities, auto-tagging, and export options, Cognio aims to be a robust backend for memory-rich AI workflows. The server exposes a REST API with endpoints for saving memories, searching, listing, exporting, and summarizing content, plus a dashboard UI for interactive browsing. The MCP tooling is exposed through 11 specialized operations, including save_memory, search_memory, list_memories, and project-scoped operations like set_active_project and get_active_project, which help you organize and isolate contexts across multiple projects. On startup, Cognio can auto-configure MCP clients and generate usage documentation in cognio.md for your workspace. You can access the web UI at /ui and the API docs at /docs, with a separate interactive docs experience available at /docs. Use the provided 11 tools to manage memories, search results, and project contexts, enabling semantic retrieval and efficient memory management for AI assistants.

Prerequisites:

• Git installed on your system
• Node.js (recommended latest LTS) if you choose to run the MCP server directly; alternatively, Docker Compose is supported as shown in the Quick Start
• Docker and Docker Compose if you prefer Docker-based deployment

Option A - Docker Compose (recommended to start quickly):

1. Clone the repository and navigate to Cognio: git clone https://github.com/0xReLogic/Cognio.git cd Cognio

2. Start the server with Docker Compose (as per Quick Start): docker-compose up -d

3. Open the UI at http://localhost:8080/ui and API docs at http://localhost:8080/docs

Option B - Run Node MCP server directly:

1. Install dependencies for the MCP server: cd Cognio cd mcp-server npm install

2. Start the MCP server (adjust environment variables as needed): NODE_ENV=production API_HOST=0.0.0.0 API_PORT=8080 DB_PATH=./data/memory.db EMBED_MODEL=all-MiniLM-L6-v2 npm start

3. The MCP server will start and expose endpoints at http://0.0.0.0:8080. Use http://localhost:8080/docs for API reference and http://localhost:8080/ui for the dashboard.

Prerequisites summary: Node.js (for direct run) or Docker/Docker Compose (recommended for ease of setup). Ensure you have network access to download dependencies and, if using auto-tagging or external embeddings, provide the necessary API keys in environment variables as described in the .env.example.

Tips and common considerations:

• Environment variables mirror those in .env.example; they control embedding, search behavior, auto-tagging, and summarization. Copy .env.example to .env and customize as needed.
• If you enable LEANN or advanced summarization, ensure sufficient memory and compute, as these features can be resource-intensive.
• The Active Project workflow helps to isolate memories per project. Always set an active project when working across multiple workspaces to avoid cross-contamination.
• The MCP setup script (in mcp-server/scripts/setup-clients.js) can auto-configure support for multiple clients. Running npm run setup from the mcp-server directory will generate MCP configs for supported clients.
• If you encounter API timeouts or startup errors, check the Docker logs or server logs at the API host/port you configured. Validate that the database path exists and is writable.
• The REST API supports exporting memories to JSON or Markdown; use /memory/export for data portability or backup.
• The Cognio UI will auto-detect the API server; it will adapt to localhost, Docker, or remote deployments. Ensure you expose the API host/port correctly in your environment.