What is Babber?

A full-stack platform for
autonomous AI agents

Babber is an open-source, self-hosted AI agent platform. Create intelligent agents with custom personalities, belief systems, and goals — then let them run autonomously using local LLMs via Ollama or any OpenAI-compatible API. All your data stays on your machine.

🧠

Intelligent Agents

Agents with personalities, beliefs, aspirations, and long-term memory. Each agent is a unique entity that evolves over time.

🔄

Autonomous Execution

Agents run autonomously using thinking protocols — making decisions, executing skills, and tracking their own TODO lists.

🔒

Fully Self-Hosted

Everything runs on your hardware. MongoDB, Redis, ChromaDB, Ollama — all in Docker. Zero cloud dependencies.

🔌

Plugin System

Extend with addons that bundle backend routes, frontend pages, skills, and protocols in a single package.

Features

Everything you need to
run an AI workforce

Core

Agent System with Beliefs & Aspirations

Create agents with custom personalities, system prompts, and per-agent generation parameters. Each agent has a belief system with immutable core beliefs and mutable additional beliefs, plus aspirations — dreams, desires, and concrete goals with lock/unlock boundaries.

Core beliefs (immutable) + Additional beliefs (mutable)
Dreams, desires, goals with priority & deadline tracking
Per-agent facts, events, and analysis topics
Multi-model support — assign different LLMs per role
Self-thinking mode for internal reasoning

agent-config.json

{
  "name": "Research Analyst",
  "personality": "Meticulous, curious, data-driven",
  "beliefs": {
    "core": ["Always verify sources", "Data over opinions"],
    "additional": ["Prefer academic papers"]
  },
  "aspirations": {
    "goals": ["Master market analysis"],
    "dreams": ["Build predictive models"]
  },
  "model": "llama3.1:70b",
  "temperature": 0.4,
  "autonomous": true
}

Reasoning

Thinking Protocols

Define step-by-step reasoning workflows for your agents. Standard protocols for structured analysis, orchestrator protocols that delegate to child protocols, and loop protocols for autonomous work cycles.

Visual flow editor with drag-and-drop step builder
Standard, Orchestrator, and Loop protocol types
Step types: action, loop, decision, delegate, todo
Response style presets: humanized, formal, technical, creative
Protocol duplication and sharing between agents

▶ Start

↓

📋 Analyze Context

↓

🔀 Decision Point

↙

🔍 Research

↘

🤝 Delegate

↓

✍️ Generate Report

↓

✓ Complete

Interface

Multi-Model & Multi-Agent Chat

A VS Code-style resizable chat panel with persistent sessions. Compare responses from different models side by side, or have multiple agents collaborate in a single conversation.

Persistent chat sessions with full history
Multi-model mode — compare LLM responses
Multi-agent mode — agents collaborate in chat
Auto-generated titles, unread badges, session search
Markdown rendering with syntax highlighting
Conversation summarization for long sessions

Research Analyst — llama3.1:70b

Analyze the latest AI agent frameworks

Research Analyst llama3.1:70b

Based on my analysis of current frameworks, I've identified three key trends... ▊

Autonomous

Autonomous Agent Execution

Let agents run independently — processing tasks, making decisions, invoking skills, and writing code. Two execution modes: continuous (until stopped) or cycle-based (N iterations).

Loop protocols for self-directed work cycles
TODO list generation — agent plans its own work
Autonomous skill invocation and parsing
Cycle tracking: tokens, duration, summaries
Start/stop controls with real-time status

● Running Autonomous Mode — Cycle 7/∞

✓ Analyzed pending tasks

✓ Invoked web_search skill

✓ Extracted 3 new facts

⟳ Writing analysis report...

○ Update project status

Personal

Creator Profile

A rich personal profile that agents use to personalize their behavior. Your identity, strengths, goals with sub-goals, dreams, cities of interest — everything agents need to understand you.

Identity, strengths, principles & values
Hierarchical goals with sub-goals, priorities, deadlines
Dreams and ideas tracking
Cities of interest for weather and travel context
Context toggles — control what agents see

Identity

Full-stack developer & AI researcher

Active Goals

HIGH Launch AI agent platform Mar 2026

✓ Backend API complete

⟳ Landing page design

Cities

🏠 Berlin ✈️ Lisbon 📍 Tokyo

Integration

Messenger & Telegram Integration

Connect agents to Telegram with full auth flow, trusted user permissions, humanized responses, and typing indicators. Your agent becomes a real chat companion.

Telegram via Telethon with 2FA support
Trusted users with granular permissions
Humanized response mode with casual tone
Typing indicator simulation
Per-account context limits and response delays
Message history, stats, and logs

🤖

Research Analyst

online

Hey, what's the latest on the AI agents market?

typing...

I've been tracking this closely! Three main trends are emerging: local LLM adoption is accelerating, multi-agent systems are becoming more practical, and...

Architecture

Built on a
modern stack

Everything runs in Docker. Backend and frontend can run locally for development with hot reload.

Vue 3 + Vuetify 3

Modern reactive frontend with Composition API, Pinia stores, and Material Design 3

Frontend

FastAPI + Pydantic

High-performance async Python backend with automatic validation and OpenAPI docs

Backend

MongoDB 7

All data stored in MongoDB via Motor async driver. Flexible schema for agents, chats, skills, and more

Database

Ollama + OpenAI

Run models locally with Ollama or connect any OpenAI-compatible API — GPT-4, Claude, Mistral

LLM

ChromaDB

Vector database for semantic long-term memory. Agents search and connect their own memories

Memory

Redis

In-memory cache for fast session storage, rate limiting, and real-time pub/sub communication

Cache

Default Ports

4200Frontend

4700Backend API

4717MongoDB

4379Redis

4800ChromaDB

11434Ollama

Skills

60+ built-in skills
for any task

Agents invoke skills autonomously. Each skill is a modular tool that an agent can call during conversation or autonomous execution.

🌐 Web

web_searchweb_fetchweb_scraperss_read

📁 Files

file_readfile_writepdf_read

💻 Code

code_executecode_reviewshell_execgit_operationsdocker_manage

🧠 Memory

memory_storememory_searchmemory_deep_processrecall_knowledge

📊 Data

csv_parsejson_parsexml_parseyaml_parseregex_extract

🎨 AI & Media

image_analyzeimage_generatesound_generatespeech_recognizevideo_watchhumanize_text

📋 Knowledge

fact_extractfact_saveevent_savestudy_materialcreator_context

🚀 Project

project_file_readproject_file_writeproject_run_codeproject_search_codeproject_context_build

📨 Communication

email_sendtelegram_sendnotification_sendask_agent

And More

Packed with
powerful tools

🎬

Video Transcription

Add videos from YouTube, TikTok, Instagram, Facebook, Twitter. Auto-transcribe and discuss with agents.

📚

Research Resources

Manage trusted sources with trust levels. Simple and deep research modes for different depth needs.

📝

Notes & Ideas

Full note and idea management with categories, priorities, drag-and-drop reordering, and AI suggestions.

🔐

VPN & Proxy

Built-in proxy management with bulk import, auto-testing, country filtering. Route agent traffic privately.

💾

Backup System

Manual and auto-backups with configurable intervals. Download, restore, or clean up old backups.

✂️

Text Selection Popup

Select any text in the app and instantly save it as a note, goal, fact, belief, or analysis topic.

📊

Analysis Topics

Create structured analysis topics, connect facts, and discuss findings with agents. Per-agent or global.

🎤

Audio TTS & STT

Text-to-speech with multiple voices and speech-to-text recognition. Per-agent voice selection with preview.

📂

Project Management

Create projects with file browsing, task management, and code editing. Agents write code to isolated dirs.

Addons

Extend with
plugins

Addons bundle backend routes, frontend pages, skills, and protocols into a single installable package.

💰

Budgeting

Personal budget management: income, expenses, accounts, loans with monthly history and agent integration.

Skills: budget_view budget_add_entry

📈

Polymarket

Prediction market integration: view events, odds, place bets and track betting history via Polymarket API.

Skills: polymarket_events polymarket_place_bet

View All Addons →

Get Started

Up and running
in 3 commands

See the full Installation Guide for detailed setup on macOS, Linux, and Windows.

1

Clone & Install

git clone https://github.com/okhmat-anton/babber.git
cd babber && make install

Copies config, starts Docker services, pulls the default Ollama model.

2

Run

make run

Starts all services in production mode. For dev mode with hot reload: make run-dev

3

Open & Create

http://localhost:4200

Login with admin / admin123, create your first agent, and start chatting.

Your Personal AI Agent Army

A full-stack platform forautonomous AI agents