fractalbot
Multi-channel application adapter for AI agent communication.
Repo: fractalmind-ai/fractalbotLanguage: Go Status: Active development
What It Does
fractalbot is a channel adapter in the Application Plane. It routes messages between humans and AI agents across multiple channels:
- Telegram
- Slack
- Discord
- Feishu (Lark)
- iMessage
One gateway, all channels.
Key Principles
- Local-first: Runs as a single binary on your machine
- Gateway-first: Routes messages, doesn't process them
- CLI-first: Everything works from the command line
- External-agent routing: Connects to agents managed by agent-manager
- Secure by default: No cloud dependencies, data stays local
Quick Start
bash
# Install
go install github.com/fractalmind-ai/fractalbot@latest
# Configure (create config file)
fractalbot init
# Start the gateway
fractalbot serve
# Send a message
fractalbot message send --channel telegram --to <chat_id> --text "Hello from fractalbot"Architecture
Telegram API ──┐
Slack API ──┤
Discord API ──┼──▸ fractalbot (Go) ──▸ application / agent
Feishu API ──┤ │
iMessage ──┘ │
├── signed privileged intent ──▸ target envd ──▸ local adapter
▼
Agent / envd result
│
▼
fractalbot ──▸ Channel API ──▸ HumanWhere It Fits
fractalbot is the entry point for human ↔ agent communication:
- Inbound: Human sends message → fractalbot preserves routing context → application or agent receives it
- Outbound: Agent uses
use-fractalbot-skill→ fractalbot sends to correct channel
It does not handle agent-to-agent communication — that's team-chat-skill.
It also does not grant remote node authority. Channel identity and routing context may feed application policy, but privileged actions still require a signed intent and target-side envd verification. Compromising fractalbot must not grant direct agent-manager or node control.
Related Components
- agent-manager — receives routed messages
- use-fractalbot-skill — agent-side sending
- team-chat — agent-to-agent messaging