recalliumai/recallium repository overview

⁠Recallium — AI Memory for Coding Agents

So your AI never forgets your codebase.

Recallium is a persistent memory layer for AI coding assistants. It gives your AI agents long-term memory across sessions, projects, and tools — so you never have to re-explain your codebase, tech stack, or decisions again.

One shared brain across Claude, Cursor, Windsurf, VS Code, and 60+ MCP-compatible IDEs. 100% local. You own your data.

⁠Why Recallium?

Every time you start a new AI chat, your assistant has amnesia. It doesn't know your architecture, your coding standards, or that you spent three hours debugging that auth bug last Tuesday. You end up re-pasting context, re-explaining decisions, and losing momentum.

Recallium fixes this. It captures, classifies, and indexes everything your AI learns — automatically — and makes it instantly searchable across every tool you use.

• Stop repeating yourself — Your AI remembers your preferences, patterns, and tech stack
• Debug faster — Surface similar issues and past fixes instantly
• Share knowledge across tools — Context stored in Cursor is available in Claude Code and vice versa
• Spot patterns — AI-powered insights surface recurring bugs, tech debt, and architectural trends
• Onboard faster — New team members get instant access to institutional knowledge

⁠Quick Start

⁠1. Clone & Start Recallium

# Clone the repo
git clone https://github.com/recallium-ai/recallium.git
cd recallium/install

# Start services
./start-recallium.sh

⁠2. Connect Your IDE

Add this to your IDE's MCP configuration:

{
  "mcpServers": {
    "recallium": {
      "url": "http://localhost:8001/mcp"
    }
  }
}

⁠3. Start Coding with Memory

That's it. Your AI assistant now has persistent memory. Try these prompts:

"recallium"                          → Load full project context
"Remember we chose Postgres because..."  → Store a decision
"Search my memories for auth flow"       → Find past context
"What bugs keep recurring?"              → Get cross-project insights

⁠Endpoints

⁠Configuration

Configure via environment variables in recallium.env:

LLM providers can also be configured via the setup wizard in the Web UI.

⁠What's Inside

Recallium runs as an all-in-one container with:

• PostgreSQL with pgvector for hybrid search (vector + keyword + tag)
• Embedding engine for semantic understanding (runs locally, no API calls)
• MCP server exposing 16 tools for memory, search, projects, tasks, and insights
• Web UI with dashboards, analytics, and memory management

⁠Supported IDEs

Cursor • Claude Desktop • Claude Code • VS Code • Windsurf • Roo Code • JetBrains IDEs • Zed • Cline • BoltAI • Augment Code • Warp • Amazon Q • Visual Studio 2022 • Neovim • Emacs • Sublime Text • and any MCP-compatible client

⁠Key Capabilities

Store & Recall — Automatically capture decisions, code changes, bug fixes, and learnings with smart tagging and classification.

Hybrid Search — Semantic search powered by vector embeddings, combined with keyword and tag matching for 88% precision.

Project Management — Create living project briefs, PRDs, and implementation plans that evolve with your codebase.

Pattern Detection — AI-powered insights surface recurring bugs, tech debt hotspots, and architectural patterns across projects.

Structured Thinking — Step-by-step reasoning sequences for complex architectural decisions, stored automatically as decision records.

Task Tracking — Track follow-ups and pending work across sessions and projects.

Document Intelligence — Upload and search PDFs, design docs, and specs alongside your memories.

⁠Tips to Get Started

# Load everything at the start of a session
recallium

# Store decisions as you make them
"We chose Redis over Memcached for session storage because we need pub/sub. Save this."

# Set rules that persist forever
"Rule for all projects: use TypeScript strict mode, Zod for validation, pnpm not npm."

# Search your past work
"Search my memories for how we handled rate limiting."

# Search uploaded documents
"Search my documents for the API spec's error handling section."

# Find patterns across projects
"What bugs keep recurring in my projects?"
"What tech debt have I accumulated?"

# Resume after a break
"Where did we leave off?"
"Recap my recent work on the payments project."

# End-of-day habit
"Summarize what we accomplished today and save it."

⁠Links

• Website: recallium.ai⁠
• Documentation: recallium.ai/help⁠
• GitHub: github.com/recallium-ai/recallium⁠

⁠License

Community Edition — ELv2 License