Like Cursor + Claude Code + Zo, but runs 100% on your machine.
- OpenAI — GPT-4o, GPT-4-turbo, GPT-3.5-turbo
- Anthropic — Claude Sonnet, Claude Opus
- Groq — Llama 3.3 70B, Mixtral (FREE tier available)
- Ollama — Run completely offline with local models
| Tool | What it does |
|---|---|
read_file |
Read any file (text, PDF, DOCX, images) |
write_file |
Create/overwrite files |
edit_file |
Surgical edits |
list_directory |
Browse folders |
search_files |
Find by name or content |
delete_file |
Remove files/folders |
run_command |
Execute shell commands |
analyze_code |
AST analysis, complexity, security |
git_command |
Git operations |
web_fetch |
HTTP requests |
web_search |
DuckDuckGo search |
remember |
Store in long-term memory |
recall |
Retrieve from memory |
- SQLite database (~/.sentience/sentience.db)
- Conversations, messages, memory
- LZ4 compression for large content
- Knowledge graph for entities
- Load custom Python skills
- Persistent across sessions
- Auto-registration
- Schedule background tasks
- Run autonomous workflows
- Download
Sentience-v2.zipfrom Releases - Extract to any folder
- Run
install.batas Administrator - Launch: Open terminal and run
sentience
Or manually:
pip install -r requirements.txt
python cli.pypip install -r requirements.txt
python cli.pyFirst run will ask for your API key, or set via environment:
# OpenAI
export OPENAI_API_KEY=sk-...
# Anthropic
export ANTHROPIC_API_KEY=sk-ant-...
# Groq (FREE tier at console.groq.com)
export GROQ_API_KEY=gsk_...Sentience v2.0 - Local AI Computer
You: read the file src/main.py and explain what it does
Sentience: I'll read the file and analyze it...
[uses read_file tool]
You: refactor it to use async
Sentience: [uses edit_file tool]
You: commit the changes
Sentience: [uses git_command tool]
| Command | Description |
|---|---|
help |
Show all commands |
new |
Start new conversation |
list |
List conversations |
load <id> |
Load conversation |
provider |
List available providers |
use <p> [m] |
Switch provider/model |
tools |
List all tools |
config |
Show configuration |
quit |
Exit |
You: analyze the code in src/api.py for security issues
Sentience: I'll analyze the code...
[uses analyze_code tool]
Found 3 security warnings:
- Line 45: Use of eval() is dangerous
- Line 78: shell=True can be dangerous
- Line 102: Pickle can execute arbitrary code
You: create a new React component called Dashboard in src/components/
Sentience: I'll create the file...
[uses write_file tool]
Created src/components/Dashboard.jsx
You: what's the latest news about AI agents?
Sentience: [uses web_search tool]
Here are the latest results:
1. "OpenAI releases new agent framework..."
2. "Anthropic announces Claude agent capabilities..."
sentience-v2/
├── core/
│ ├── engine.py # Main agent loop
│ ├── config.py # Configuration & BYOK
│ ├── memory.py # SQLite + compression + graph
│ └── tools.py # 15+ real tool implementations
├── cli.py # Terminal interface
├── requirements.txt # Dependencies
├── install.bat # Windows installer
└── README.md
| Feature | Sentience | Cursor | Claude Code | Zo |
|---|---|---|---|---|
| 100% Local | ✅ | ❌ | ❌ | ❌ |
| BYOK | ✅ | ❌ | ❌ | ✅ |
| Offline | ✅ | ❌ | ❌ | ❌ |
| Free Tier | ✅ (Groq) | ❌ | ❌ | ❌ |
| Open Source | ✅ | ❌ | ❌ | ❌ |
| Shell Access | ✅ | ❌ | ✅ | ✅ |
| Memory | ✅ | ❌ | ❌ | ✅ |
| Knowledge Graph | ✅ | ❌ | ❌ | ✅ |
- Python 3.11+
- 4GB RAM minimum
- Internet for API calls (or use Ollama for offline)
MIT
Inspired by:
- Zo Computer — architecture
- Claude Code — tool calling
- Cursor — coding assistant