cachibot 0.2.26.dev1

The Armored AI Agent. Cross-platform, secure, yours.

CachiBot

The Armored AI Agent

Visual. Transparent. Secure.

Español · 中文版 · Português

A visual AI agent platform with full transparency. Named after the Venezuelan cachicamo (armadillo) — built to be armored, auditable, and yours to control.

Quick Start · Features · Architecture · Security · Contributing · Discord

---

Why Visual?

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

CLI-based agents operate in a black box — no visibility into running tasks, no way to monitor multiple bots, no real-time insight into what the agent is doing.

CachiBot gives you full visibility. Watch your bots work through a dashboard, see every task and job in a clean interface, approve or reject actions before they execute, and maintain a 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 WebSocket 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 isolation with AST-based risk analysis
• Approval Flow — Visual approval for risky operations before they execute
• Multi-Provider — Kimi K2.5, Claude, OpenAI, Ollama, Groq, and more

Quick Start

1. Install

One-line install (Linux / macOS — installs Python, venv, and service):

curl -fsSL https://cachibot.com/install.sh | bash

Or install with pip (if you already have Python 3.10+):

pip install cachibot

2. Set your API key

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

# Or Claude
export ANTHROPIC_API_KEY="your-key"

# Or OpenAI
export OPENAI_API_KEY="your-key"

3. Launch

cachibot server

Open http://localhost:6392 — the frontend is bundled and served automatically.

CLI Usage

cachibot server                              # Start the dashboard
cachibot "list all Python files"             # Run a single task
cachibot                                     # Interactive mode
cachibot --model anthropic/claude-sonnet-4-20250514 "explain this"  # Specific model
cachi server                                 # Short alias

Architecture

graph TB
    subgraph Frontend["React Dashboard"]
        Bots[Bots]
        Chats[Chats]
        Jobs[Jobs & Tasks]
        KB[Knowledge Base]
        Conn[Connections]
    end

    subgraph Backend["FastAPI Backend"]
        Agent["Prompture Agent"]
        Tools["Tool Registry"]
        Sandbox["Sandbox Executor"]
    end

    subgraph Providers["LLM Providers"]
        Moonshot[Moonshot/Kimi]
        Claude[Claude]
        OpenAI[OpenAI]
        Ollama[Ollama]
        Groq[Groq]
    end

    subgraph Platforms["Platform Connections"]
        Telegram[Telegram]
        Discord[Discord]
    end

    Frontend -- "WebSocket / REST" --> Backend
    Backend --> Providers
    Backend --> Platforms

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 needed)
Groq	groq/llama-3.1-70b	GROQ_API_KEY

Security

CachiBot is built with security as a core principle. Visibility is security — the biggest risk with AI agents is not knowing what they're doing.

Sandboxed Execution

Python code runs in a restricted environment:

• Import Restrictions — Only safe modules allowed (json, math, datetime, etc.)
• Path Restrictions — File access limited to 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, importlib, eval, exec, pickle, marshal.

Roadmap

• Visual dashboard with real-time monitoring
• Multi-bot management
• Sandboxed Python execution
• Multi-provider LLM support
• Knowledge base with document upload
• Telegram integration
• Discord integration
• Plugin marketplace
• Voice interface
• Mobile companion app

Contributing

Contributions are welcome!

git clone https://github.com/jhd3197/CachiBot.git
cd CachiBot

# Backend
pip install -e ".[dev]"
cachibot server --reload

# Frontend (in another terminal)
cd frontend && npm install && npm run dev

# Tests & linting
pytest
ruff check src/
cd frontend && npm run lint

Community

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