CLIver

A Personal AI Lab — an embeddable Python agent engine, with a CLI application in progress.

Get Started Learn More

What CLIver Can Do

🔌

Model-Agnostic

Connect to any OpenAI-compatible API, Anthropic-protocol endpoints, or local models via Ollama. Switch providers on the fly — no vendor lock-in.

🧠

AgentCore Engine

A clean Re-Act loop with chat() and stream() APIs. Supports text, image, audio, and video generation when configured. Embed into your own Python applications.

🛠️

Built-in Tools

Nine tools for file I/O, shell execution, code, web fetching, and todo tracking — all sharing one permission and event pipeline.

🎯

Skills System

LLM-driven skill activation following the Agent Skills specification. Built-in skills for brainstorming, planning, and structured execution.

🔐

Opt-in Policy

SAFE/ASK permission gates with an application-supplied callback, plus execution scopes for environment, working directory, and paths.

⚡

MCP Support

Connect stdio and streamable HTTP servers. MCP tools share the same execution path as builtin tools; prompts and resources are read on demand.

Quick Start

Install the core library and run your first agent.

pip install cliver-core==0.3.0a5

python - <<'PY'
import asyncio
from cliver_core import AgentCore, ModelBindings, presets

agent = AgentCore(ModelBindings(llm=presets.deepseek()))
print(asyncio.run(agent.chat("Hello!")).message.content)
PY