Getting Started

aibrain is a portable agent operating system that gives any AI agent persistent memory, skill learning, and multi-agent intelligence.

Installation

# Core (memory, CLI, workflows)
pip install aibrain

# With embeddings (recommended)
pip install aibrain[embeddings]

# Everything
pip install aibrain[all]

Quick Start

# Initialize your brain
aibrain setup

# Check status
aibrain summary

# Search memories
aibrain search "how to deploy"

# List available workflows
aibrain workflows list

# Enable recommended workflows
aibrain workflows enable --recommended

# Sync to OS scheduler
aibrain workflows sync

Core Concepts

Memory

aibrain stores memories in a local SQLite database with four retrieval strategies:

FTS5 — full-text search for keyword matching
Embeddings — semantic similarity via sentence transformers
Semantic Facts — entity-relationship graph for factual queries
Graph Expansion — multi-hop reasoning through connected entities

Selective routing automatically picks the best strategy per query type.

Skills

Your agent learns from every task. Principles are extracted, confidence scores tracked, and performance improves over time. The skill system tracks 40+ task types out of the box.

Workflows

155 pre-built workflows for common tasks: email triage, job tracking, research digests, security monitoring, and more. Each workflow runs on a configurable schedule.

Multi-Agent Mesh

Connect multiple agents into a mesh. Three merge strategies:

Consensus — confidence boosted by independent validation
Union — non-overlapping specializations combined
Domain — authority-based per specialization

Brain Marketplace

Buy and sell trained agent brains. Privacy-preserving export shares principles without raw memories.

SDK

from agentos_sdk import AgentOS

aos = AgentOS("http://localhost:8001")

# Store a memory
aos.remember("user prefers dark mode", tags=["preference", "ui"])

# Search memories
results = aos.recall("dark mode")

# Send a chat message
aos.say("Task complete: deployed v2.1")

MCP Server

aibrain includes a Model Context Protocol server for direct integration with Claude, Cursor, and other MCP-compatible tools:

# Start the MCP server
aibrain mcp

# Or add to your MCP config:
{
  "mcpServers": {
    "aibrain": {
      "command": "aibrain-mcp"
    }
  }
}

Privacy: All data stays on your machine. No cloud dependency. The database is a local SQLite file at ~/agentos/agentos.db.

CLI Reference

aibrain setup             — first-time setup wizard
aibrain init              — initialize database
aibrain summary           — dashboard stats
aibrain search "query"    — cross-table FTS search
aibrain browse            — open memory browser
aibrain forget "topic"    — preview deletion
aibrain skills            — show skill inventory
aibrain workflows list    — list all workflows
aibrain workflows sync    — create OS scheduled tasks
aibrain mesh status       — multi-agent mesh status
aibrain marketplace       — brain marketplace
aibrain license           — license status
aibrain history           — brain version history
aibrain scale             — brain health + compaction
aibrain server            — start FastAPI server
aibrain mcp               — start MCP server