MAMA
Always-on companion for Claude that remembers your decisions and their evolution. Local-first memory using SQLite + transformers.js embeddings.
claude mcp add --transport stdio jungjaehoon-lifegamez-mama npx -y @jungjaehoon/mama-os
How to use
MAMA is a 24/7 autonomous AI agent platform that runs on your machine and connects to messaging platforms like Discord, Slack, and Telegram. It uses a modular skill system where you can drop .md skill files into ~/.mama/skills/ to teach the agent new capabilities, or install additional MCP servers and plugins to extend its workflow orchestration. The system includes a built-in cron scheduler so tasks can run on a schedule, and a web dashboard for managing skills, cron jobs, and agent behavior. Users interact with MAMA through natural language prompts, and MAMA will route tasks to the appropriate skills, remember decisions with reasoning, and report outcomes back to you through the connected channels.
Key capabilities include:
- Always-on agent loop with multi-platform presence (Discord/Slack/Telegram).
- Installable skills via simple Markdown files that describe how the agent should operate.
- Built-in cron scheduler to automate periodic tasks and workflows.
- A Conductor and multi-agent system for delegation, dynamic workflows, and coordinated decision-making.
- Integration with external backends (via CLI subprocesses like Claude/Codex) for advanced tasks while preserving stability.
To use, install the provider package (MAMA OS), initialize the local skill store, and start the daemon. You can then configure your messaging platform bots, add skills, and create cron jobs that trigger skills or workflows. The dashboard will provide visibility into scheduled tasks and agent decisions.
How to install
Prerequisites:
- Node.js 22.x or newer installed on your machine
- npm (comes with Node.js)
- Optional: Docker for secured deployments (recommended by the project)
Step-by-step:
-
Install the MAMA OS CLI globally (as shown in the README): npm install -g @jungjaehoon/mama-os
-
Initialize the skill directory and start the dashboard/server: mama init mama start
This will copy default skills to ~/.mama/skills/ and launch the web dashboard at localhost:3847.
-
Connect messaging platforms:
- Create and configure Discord/Slack/Telegram bots as per platform documentation.
- In the MAMA dashboard, add the corresponding bot integrations and authorize them to your workspace/server.
-
Add skills:
- Create Markdown skill files in ~/.mama/skills/ describing the prompts and actions.
- Reload or restart the MAMA OS daemon to pick up new skills.
-
Optional: Run MAMA OS inside Docker for isolation: docker run -it --rm -p 3847:3847 -v ~/.mama:/root/.mama @jungjaehoon/mama-os
-
Verify:
- Open http://localhost:3847 to view the dashboard.
- Ensure that your bots show as online and cron jobs are visible in the dashboard.
Additional notes
Tips and common issues:
- Security: MAMA OS runs autonomous agents with filesystem access. Run in a restricted environment (Docker, isolated VM, or sandbox) when possible.
- Skill formats: Markdown-based skills are simple to author but ensure they include clear instructions, prompts, and fallback behaviors.
- Cron reliability: Ensure the host system clock is correct and that the daemon has permissions to write to its state directory.
- CLI subprocess: When using Claude/Codex backends via CLI, ensure the respective CLIs are installed and authenticated to avoid runtime errors.
- Backups: Regularly back up ~/.mama/skills and configuration so you can recover quickly after a disk failure.
- Security posture: Review the Security Guide and Standalone Compliance Notes referenced in the project README for deployment considerations and policy settings.
Related MCP Servers
mcp-memory-keeper
MCP server for persistent context management in AI coding assistants
sharedcontext
MCP server that gives AI coding assistants persistent cross-client memory. Facts and conversations stored in SQLite, encrypted with AES-256-GCM, synced to Arweave. No server, no account, recoverable with a 12-word phrase.
code-memory
MCP server with local vector search for your codebase. Smart indexing, semantic search, Git history — all offline.
mcp-ragex
MCP server for intelligent code search: semantic (RAG), symbolic (tree-sitter), and regex (ripgrep) search modes. Built for Claude Code and AI coding assistants.
second-brain-health-check
Context engineering health check for Claude Code. Scores CLAUDE.md, skills, hooks, and memory against 7 CE patterns — tells you exactly what to fix. Free MCP, runs in 60 seconds.
mem0
🧠 Production-ready MCP server providing intelligent memory for Claude Code with async architecture, Neo4j knowledge graphs, smart chunking & enterprise security. One-command Docker deployment.