recallium

Recallium acts as a persistent memory layer for your AI agents, providing a centralized knowledge store that can be recalled across sessions, projects, and tools. Once running, you can connect your MCP-compatible IDEs and agents to the web dashboard at http://localhost:9001 to configure providers, manage memory, and enable cross-project intelligence. Use the core features to attach project briefs, PRDs, specifications, and other artifacts as memory inputs, and then invoke the memory load by name (the magic word, recallium, or your IDE’s integration cue) to accelerate reasoning with context. The dashboard guides you through provider selection (Ollama/local models, OpenAI, Anthropic, Gemini, or OpenRouter) and enables automatic failover and memory tagging for quick retrieval. The included tools support uploading documents, linking projects, and organizing knowledge so that your AI agents remember patterns, decisions, and workflows across environments.

Key capabilities you can leverage include:

• One Memory System: A single memory backbone that works across Cursor, Claude Desktop/Code, VS Code, Windsurf, JetBrains, Zed, and other MCP-compatible tools.
• Cross-Project Intelligence: Lessons and patterns learned in one project can be applied automatically to others.
• Document Knowledge Base: Upload PDFs, docs, specs, and other materials so the AI understands your system instantly.
• Smart Project Linking: Connect v1, v2, v3 patterns and share learnings across related projects.
• Zero-Config Memory: Auto-tags, auto-clusters, and automatic learning from interactions with minimal setup.
• Privacy-First: Runs locally with data staying on your machine when you choose local models.
• Free Local Option: Use Ollama + built-in embeddings to avoid API costs when possible.

Prerequisites:

• Docker and Docker Compose installed on your machine
• Basic familiarity with running containerized services
• Optional: Ollama or local embeddings if you want a fully local setup

Install steps:

1. Clone the repository or download the Recallium package.
2. Navigate to the install directory and ensure the recallium.env file is present with appropriate configuration.
3. Start Recallium using Docker Compose:

cd install
docker compose --env-file recallium.env pull
docker compose --env-file recallium.env up -d

4. Open the web dashboard to complete setup at http://localhost:9001. Follow the wizard to configure your LLM provider and memory settings.
5. (Optional) If you prefer one-click scripts on macOS/Linux/Windows, run the provided scripts:

• macOS/Linux: ./start-recallium.sh or ./start-recallium.sh
• Windows: use the recallium.bat script after any necessary Ollama/Docker prerequisites

cd install
# Start scripts
./start-recallium.sh

Prerequisites note: If you are using Docker on Linux with IPv6, ensure IPv6 is enabled in Docker if you encounter port binding issues. See the Quick Start notes in the README for details.

Tips and caveats:

• If you’re using a local Ollama model or embedded embeddings, ensure Ollama connectivity is configured (OLLAMA_HOST) and accessible from the Recallium container or host.
• When using Docker Compose, you can customize recallium.env to set memory limits, provider priorities, and failover behavior. See recallium.env for available options.
• For macOS/Linux, the default setup uses IPv6 dual-stack binding. If you experience port binding issues, you can modify the start scripts to bind to 0.0.0.0 instead of [::].
• Remember the magic word recallium in IDEs to trigger loading of project context and memory when connected to Recallium.
• If you want to switch providers later, use the web dashboard setup wizard to reconfigure LLM providers and failover without losing memories.