CachiBot

The Armored AI Agent

Visual. Transparent. Secure.

---

CachiBot is a visual AI agent platform with full transparency. Named after the Venezuelan cachicamo (armadillo), it's built to be armored, auditable, and yours to control.

Why Visual?

Most AI agent tools run in terminals where you can't see what's happening. That's a security nightmare.

The problem with CLI-based agents:

• You can't see what the agent is doing in real-time
• No visibility into running tasks or jobs
• No way to monitor multiple bots simultaneously
• Actions happen in a black box

CachiBot gives you full visibility:

• Watch your bots work in real-time through the dashboard
• See every task, job, and chat in a clean interface
• Monitor connections to Telegram, Discord, and other platforms
• Approve or reject actions before they execute
• Full audit trail of everything your bots do

Features

• Visual Dashboard — See all your bots, their status, and activity at a glance
• Real-time Monitoring — Watch tasks and jobs execute with live updates
• Multi-Bot Management — Create and manage multiple specialized bots
• Platform Connections — Connect bots to Telegram, Discord, and more
• Knowledge Base — Upload documents to give bots specialized knowledge
• Secure Sandbox — Code runs in isolated environment with restricted imports
• Approval Flow — Visual approval for risky operations before they execute
• Multi-Provider — Kimi K2.5, Claude, OpenAI, and more

Quick Start

1. Install

pip install cachibot

2. Set your API key

# For Moonshot/Kimi (default)
export MOONSHOT_API_KEY="your-api-key"

# Or for Claude
export ANTHROPIC_API_KEY="your-api-key"

# Or for OpenAI
export OPENAI_API_KEY="your-api-key"

3. Run CachiBot

cachibot server

Open http://localhost:6392 in your browser. The frontend is bundled and served automatically.

CLI Usage

# Start the dashboard server
cachibot server

# Run a single task
cachibot "list all Python files"

# Interactive mode
cachibot

# Use a specific model
cachibot --model anthropic/claude-sonnet-4-20250514 "explain this code"

# Short alias works too
cachi server

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         CachiBot                                 │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────────────────────────────────────────────────┐    │
│  │                  React Dashboard                         │    │
│  │  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌────────┐  │    │
│  │  │   Bots   │  │  Chats   │  │   Jobs   │  │ Tasks  │  │    │
│  │  └──────────┘  └──────────┘  └──────────┘  └────────┘  │    │
│  │  ┌──────────┐  ┌──────────┐  ┌──────────────────────┐  │    │
│  │  │ Settings │  │Knowledge │  │    Connections       │  │    │
│  │  └──────────┘  └──────────┘  └──────────────────────┘  │    │
│  └────────────────────────┬────────────────────────────────┘    │
│                           │ WebSocket / REST                     │
│  ┌────────────────────────▼────────────────────────────────┐    │
│  │              FastAPI Backend                             │    │
│  │  ┌──────────┐  ┌──────────┐  ┌──────────────────────┐   │    │
│  │  │Prompture │  │  Tools   │  │   Sandbox Executor   │   │    │
│  │  │  Agent   │  │ Registry │  │  (Isolated Python)   │   │    │
│  │  └──────────┘  └──────────┘  └──────────────────────┘   │    │
│  └────────────────────────┬────────────────────────────────┘    │
│                           │                                      │
│  ┌────────────────────────▼────────────────────────────────┐    │
│  │             LLM Providers (via Prompture)                │    │
│  │   Moonshot  │  Claude  │  OpenAI  │  Ollama  │  Groq    │    │
│  └──────────────────────────────────────────────────────────┘    │
│                           │                                      │
│  ┌────────────────────────▼────────────────────────────────┐    │
│  │              Platform Connections                        │    │
│  │         Telegram  │  Discord  │  (more coming)          │    │
│  └──────────────────────────────────────────────────────────┘    │
└─────────────────────────────────────────────────────────────────┘

Security

CachiBot is built with security as a core principle:

Visibility = Security

The biggest security risk with AI agents is not knowing what they're doing. CachiBot solves this by making everything visible:

• See every action before it executes
• Approve or reject risky operations
• Full audit trail of all bot activity
• Real-time monitoring of running tasks

Sandboxed Execution

Python code runs in a sandboxed environment:

• Import Restrictions — Only safe modules allowed (json, math, datetime, etc.)
• Path Restrictions — Can only access files in the workspace
• Execution Timeout — Code killed after timeout (default: 30s)
• Risk Analysis — AST-based detection of dangerous operations

Always Blocked

These are never allowed regardless of configuration:

• subprocess, os.system, ctypes
• socket, ssl, raw network access
• importlib, eval, exec (dynamic code)
• pickle, marshal (unsafe serialization)

Supported Models

Provider	Model	Environment Variable
Moonshot	moonshot/kimi-k2.5	MOONSHOT_API_KEY
Claude	anthropic/claude-sonnet-4-20250514	ANTHROPIC_API_KEY
OpenAI	openai/gpt-4o	OPENAI_API_KEY
Ollama	ollama/llama3.1:8b	(local, no key)
Groq	groq/llama-3.1-70b	GROQ_API_KEY

Contributing

Contributions are welcome!

# Clone the repo
git clone https://github.com/jhd3197/CachiBot.git
cd CachiBot

# Install backend in development mode
pip install -e ".[dev]"

# Install frontend dependencies
cd frontend && npm install

# Start backend (dev)
cachibot server --reload

# Start frontend dev server (in another terminal)
cd frontend && npm run dev

# Run tests
pytest

# Lint
ruff check src/
cd frontend && npm run lint

License

MIT License - see LICENSE for details.

Credits

• Built with Prompture for structured LLM interaction
• Named after the Venezuelan cachicamo (armadillo)

---

Made with care by Juan Denis